Rewrite github.com to git.ipng.ch for popular repos

.drone.yml

@@ -8,9 +8,9 @@ steps:
       - git lfs install
       - git lfs pull
   - name: build
-    image: git.ipng.ch/ipng/drone-hugo:release-0.130.0
+    image: git.ipng.ch/ipng/drone-hugo:release-0.145.1
     settings:
-      hugo_version: 0.130.0
+      hugo_version: 0.145.0
       extended: true
   - name: rsync
     image: drillster/drone-rsync

config.toml

@@ -1,38 +0,0 @@
-baseURL = 'https://ipng.ch/'
-languageCode = 'en-us'
-title = "IPng Networks"
-theme = 'hugo-theme-ipng'
-
-mainSections = ["articles"]
-# disqusShortname = "example"
-paginate = 4
-
-[params]
-  author = "IPng Networks GmbH"
-  siteHeading = "IPng Networks"
-  favicon = "favicon.ico" # Adds a small icon next to the page title in a tab
-  showBlogLatest = false
-  mainSections = ["articles"]
-  showTaxonomyLinks = false
-  nBlogLatest = 14 # number of blog post om the home page
-  Paginate = 30
-  blogLatestHeading = "Latest Dabblings"
-  footer = "Copyright 2021- IPng Networks GmbH, all rights reserved"
-
-  [params.social]
-    email = "info+www@ipng.ch"
-    mastodon = "@IPngNetworks"
-    twitter = "IPngNetworks"
-    linkedin = "pimvanpelt"
-    github = "pimvanpelt"
-    instagram = "IPngNetworks"
-    rss = true
-
-[taxonomies]
-  year = "year"
-  month = "month"
-  tags = "tags"
-  categories = "categories"
-
-[permalinks]
-  articles = "/s/articles/:year/:month/:day/:slug"

content/articles/2021-08-13-vpp-2.md

@@ -89,7 +89,7 @@ lcp lcp-sync off
 ```
 
 The prep work for the rest of the interface syncer starts with this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/2d00de080bd26d80ce69441b1043de37e0326e0a)], and
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/2d00de080bd26d80ce69441b1043de37e0326e0a)], and
 for the rest of this blog post, the behavior will be in the 'on' position.
 
 ### Change interface: state
@@ -120,7 +120,7 @@ the state it was. I did notice that you can't bring up a sub-interface if its pa
 is down, which I found counterintuitive, but that's neither here nor there.
 
 All of this is to say that we have to be careful when copying state forward, because as
-this [[commit](https://github.com/pimvanpelt/lcpng/commit/7c15c84f6c4739860a85c599779c199cb9efef03)]
+this [[commit](https://git.ipng.ch/ipng/lcpng/commit/7c15c84f6c4739860a85c599779c199cb9efef03)]
 shows, issuing `set int state ... up` on an interface, won't touch its sub-interfaces in VPP, but
 the subsequent netlink message to bring the _LIP_ for that interface up, **will** update the
 children, thus desynchronising Linux and VPP: Linux will have interface **and all its
@@ -128,7 +128,7 @@ sub-interfaces** up unconditionally; VPP will have the interface up and its sub-
 whatever state they were before.
 
 To address this, a second
-[[commit](https://github.com/pimvanpelt/lcpng/commit/a3dc56c01461bdffcac8193ead654ae79225220f)] was
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/a3dc56c01461bdffcac8193ead654ae79225220f)] was
 needed. I'm not too sure I want to keep this behavior, but for now, it results in an intuitive
 end-state, which is that all interfaces states are exactly the same between Linux and VPP.
 
@@ -157,7 +157,7 @@ DBGvpp# set int state TenGigabitEthernet3/0/0 up
 ### Change interface: MTU
 
 Finally, a straight forward
-[[commit](https://github.com/pimvanpelt/lcpng/commit/39bfa1615fd1cafe5df6d8fc9d34528e8d3906e2)], or
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/39bfa1615fd1cafe5df6d8fc9d34528e8d3906e2)], or
 so I thought. When the MTU changes in VPP (with `set interface mtu packet N <int>`), there is
 callback that can be registered which copies this into the _LIP_. I did notice a specific corner
 case: In VPP, a sub-interface can have a larger MTU than its parent. In Linux, this cannot happen,
@@ -179,7 +179,7 @@ higher than that, perhaps logging an error explaining why. This means two things
 1.   Any change in VPP of a parent MTU should ensure all children are clamped to at most that.
 
 I addressed the issue in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/79a395b3c9f0dae9a23e6fbf10c5f284b1facb85)].
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/79a395b3c9f0dae9a23e6fbf10c5f284b1facb85)].
 
 ### Change interface: IP Addresses
 
@@ -199,7 +199,7 @@ VPP into the companion Linux devices:
     _LIP_ with `lcp_itf_set_interface_addr()`.
 
 This means with this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/f7e1bb951d648a63dfa27d04ded0b6261b9e39fe)], at
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/f7e1bb951d648a63dfa27d04ded0b6261b9e39fe)], at
 any time a new _LIP_ is created, the IPv4 and IPv6 address on the VPP interface are fully copied
 over by the third change, while at runtime, new addresses can be set/removed as well by the first
 and second change.

content/articles/2021-08-15-vpp-3.md

@@ -100,7 +100,7 @@ linux-cp {
 
 Based on this config, I set the startup default in `lcp_set_lcp_auto_subint()`, but I realize that
 an administrator may want to turn it on/off at runtime, too, so I add a CLI getter/setter that
-interacts with the flag in this [[commit](https://github.com/pimvanpelt/lcpng/commit/d23aab2d95aabcf24efb9f7aecaf15b513633ab7)]:
+interacts with the flag in this [[commit](https://git.ipng.ch/ipng/lcpng/commit/d23aab2d95aabcf24efb9f7aecaf15b513633ab7)]:
 
 ```
 DBGvpp# show lcp
@@ -116,11 +116,11 @@ lcp lcp-sync off
 ```
 
 The prep work for the rest of the interface syncer starts with this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/2d00de080bd26d80ce69441b1043de37e0326e0a)], and
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/2d00de080bd26d80ce69441b1043de37e0326e0a)], and
 for the rest of this blog post, the behavior will be in the 'on' position.
 
 The code for the configuration toggle is in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/934446dcd97f51c82ddf133ad45b61b3aae14b2d)].
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/934446dcd97f51c82ddf133ad45b61b3aae14b2d)].
 
 ### Auto create/delete sub-interfaces
 
@@ -145,7 +145,7 @@ I noticed that interface deletion had a bug (one that I fell victim to as well:
 remove the netlink device in the correct network namespace), which I fixed.
 
 The code for the auto create/delete and the bugfix is in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/934446dcd97f51c82ddf133ad45b61b3aae14b2d)].
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/934446dcd97f51c82ddf133ad45b61b3aae14b2d)].
 
 ### Further Work
 

content/articles/2021-08-25-vpp-4.md

@@ -154,7 +154,7 @@ For now, `lcp_nl_dispatch()` just throws the message away after logging it with
 a function that will come in very useful as I start to explore all the different Netlink message types.
 
 The code that forms the basis of our Netlink Listener lives in [[this
-commit](https://github.com/pimvanpelt/lcpng/commit/c4e3043ea143d703915239b2390c55f7b6a9b0b1)] and
+commit](https://git.ipng.ch/ipng/lcpng/commit/c4e3043ea143d703915239b2390c55f7b6a9b0b1)] and
 specifically, here I want to call out I was not the primary author, I worked off of Matt and Neale's
 awesome work in this pending [Gerrit](https://gerrit.fd.io/r/c/vpp/+/31122).
 
@@ -182,7 +182,7 @@ Linux interface VPP is not aware of. But, if I can find the _LIP_, I can convert
 add or remove the ip4/ip6 neighbor adjacency.
 
 The code for this first Netlink message handler lives in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/30bab1d3f9ab06670fbef2c7c6a658e7b77f7738)]. An
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/30bab1d3f9ab06670fbef2c7c6a658e7b77f7738)]. An
 ironic insight is that after writing the code, I don't think any of it will be necessary, because
 the interface plugin will already copy ARP and IPv6 ND packets back and forth and itself update its
 neighbor adjacency tables; but I'm leaving the code in for now.
@@ -197,7 +197,7 @@ it or remove it, and if there are no link-local addresses left, disable IPv6 on
 There's also a few multicast routes to add (notably 224.0.0.0/24 and ff00::/8, all-local-subnet).
 
 The code for IP address handling is in this
-[[commit]](https://github.com/pimvanpelt/lcpng/commit/87742b4f541d389e745f0297d134e34f17b5b485), but
+[[commit]](https://git.ipng.ch/ipng/lcpng/commit/87742b4f541d389e745f0297d134e34f17b5b485), but
 when I took it out for a spin, I noticed something curious, looking at the log lines that are
 generated for the following sequence:
 
@@ -236,7 +236,7 @@ interface and directly connected route addition/deletion is slightly different i
 So, I decide to take a little shortcut -- if an addition returns "already there", or a deletion returns
 "no such entry", I'll just consider it a successful addition and deletion respectively, saving my eyes
 from being screamed at by this red error message. I changed that in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/d63fbd8a9a612d038aa385e79a57198785d409ca)],
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/d63fbd8a9a612d038aa385e79a57198785d409ca)],
 turning this situation in a friendly green notice instead.
 
 ### Netlink: Link (existing)
@@ -267,7 +267,7 @@ To avoid this loop, I temporarily turn off `lcp-sync` just before handling a bat
 turn it back to its original state when I'm done with that.
 
 The code for all/del of existing links is in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/e604dd34784e029b41a47baa3179296d15b0632e)].
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/e604dd34784e029b41a47baa3179296d15b0632e)].
 
 ### Netlink: Link (new)
 
@@ -276,7 +276,7 @@ doesn't have a _LIP_ for, but specifically describes a VLAN interface?  Well, th
 is trying to create a new sub-interface. And supporting that operation would be super cool, so let's go!
 
 Using the earlier placeholder hint in `lcp_nl_link_add()` (see the previous
-[[commit](https://github.com/pimvanpelt/lcpng/commit/e604dd34784e029b41a47baa3179296d15b0632e)]),
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/e604dd34784e029b41a47baa3179296d15b0632e)]),
 I know that I've gotten a NEWLINK request but the Linux ifindex doesn't have a _LIP_. This could be
 because the interface is entirely foreign to VPP, for example somebody created a dummy interface or
 a VLAN sub-interface on one:
@@ -331,7 +331,7 @@ a boring `<phy>.<subid>` name.
 
 Alright, without further ado, the code for the main innovation here, the implementation of
 `lcp_nl_link_add_vlan()`, is in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/45f408865688eb7ea0cdbf23aa6f8a973be49d1a)].
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/45f408865688eb7ea0cdbf23aa6f8a973be49d1a)].
 
 ## Results
 

content/articles/2021-09-02-vpp-5.md

@@ -118,7 +118,7 @@ or Virtual Routing/Forwarding domains). So first, I need to add these:
 
 All of this code was heavily inspired by the pending [[Gerrit](https://gerrit.fd.io/r/c/vpp/+/31122)]
 but a few finishing touches were added, and wrapped up in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/7a76498277edc43beaa680e91e3a0c1787319106)].
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/7a76498277edc43beaa680e91e3a0c1787319106)].
 
 ### Deletion
 
@@ -459,7 +459,7 @@ it as 'unreachable' rather than deleting it. These are *additions* which have a
 but with an interface index of 1 (which, in Netlink, is 'lo'). This makes VPP intermittently crash, so I
 currently commented this out, while I gain better understanding. Result: blackhole/unreachable/prohibit
 specials can not be set using the plugin. Beware!
-(disabled in this [[commit](https://github.com/pimvanpelt/lcpng/commit/7c864ed099821f62c5be8cbe9ed3f4dd34000a42)]).
+(disabled in this [[commit](https://git.ipng.ch/ipng/lcpng/commit/7c864ed099821f62c5be8cbe9ed3f4dd34000a42)]).
 
 ## Credits
 

content/articles/2021-09-10-vpp-6.md

@@ -88,7 +88,7 @@ stat['/if/rx-miss'][:, 1].sum() - returns the sum of packet counters for
 ```
 
 Alright, so let's grab that file and refactor it into a small library for me to use, I do
-this in [[this commit](https://github.com/pimvanpelt/vpp-snmp-agent/commit/51eee915bf0f6267911da596b41a4475feaf212e)].
+this in [[this commit](https://git.ipng.ch/ipng/vpp-snmp-agent/commit/51eee915bf0f6267911da596b41a4475feaf212e)].
 
 ### VPP's API
 
@@ -159,7 +159,7 @@ idx=19 name=tap4 mac=02:fe:17:06:fc:af mtu=9000 flags=3
 
 So I added a little abstration with some error handling and one main function
 to return interfaces as a Python dictionary of those `sw_interface_details`
-tuples in [[this commit](https://github.com/pimvanpelt/vpp-snmp-agent/commit/51eee915bf0f6267911da596b41a4475feaf212e)].
+tuples in [[this commit](https://git.ipng.ch/ipng/vpp-snmp-agent/commit/51eee915bf0f6267911da596b41a4475feaf212e)].
 
 ### AgentX
 
@@ -207,9 +207,9 @@ once asked with `GetPDU` or `GetNextPDU` requests, by issuing a corresponding `R
 to the SNMP server -- it takes care of all the rest!
 
 The resulting code is in [[this
-commit](https://github.com/pimvanpelt/vpp-snmp-agent/commit/8c9c1e2b4aa1d40a981f17581f92bba133dd2c29)]
+commit](https://git.ipng.ch/ipng/vpp-snmp-agent/commit/8c9c1e2b4aa1d40a981f17581f92bba133dd2c29)]
 but you can also check out the whole thing on
-[[Github](https://github.com/pimvanpelt/vpp-snmp-agent)].
+[[Github](https://git.ipng.ch/ipng/vpp-snmp-agent)].
 
 ### Building
 

content/articles/2021-09-21-vpp-7.md

@@ -480,7 +480,7 @@ is to say, those packets which were destined to any IP address configured on the
 plane. Any traffic going _through_ VPP will never be seen by Linux! So, I'll have to be
 clever and count this traffic by polling VPP instead. This was the topic of my previous
 [VPP Part 6]({{< ref "2021-09-10-vpp-6" >}}) about the SNMP Agent. All of that code
-was released to [Github](https://github.com/pimvanpelt/vpp-snmp-agent), notably there's
+was released to [Github](https://git.ipng.ch/ipng/vpp-snmp-agent), notably there's
 a hint there for an `snmpd-dataplane.service` and a `vpp-snmp-agent.service`, including
 the compiled binary that reads from VPP and feeds this to SNMP.
 

content/articles/2021-12-23-vpp-playground.md

@@ -62,7 +62,7 @@ plugins:
     or route, or the system receiving ARP or IPv6 neighbor request/reply from neighbors), and applying
     these events to the VPP dataplane.
 
-I've published the code on [Github](https://github.com/pimvanpelt/lcpng/) and I am targeting a release
+I've published the code on [Github](https://git.ipng.ch/ipng/lcpng/) and I am targeting a release
 in upstream VPP, hoping to make the upcoming 22.02 release in February 2022. I have a lot of ground to
 cover, but I will note that the plugin has been running in production in [AS8298]({{< ref "2021-02-27-network" >}})
 since Sep'21 and no crashes related to LinuxCP have been observed.
@@ -195,7 +195,7 @@ So grab a cup of tea, while we let Rhino stretch its legs, ehh, CPUs ...
 pim@rhino:~$ mkdir -p ~/src
 pim@rhino:~$ cd ~/src
 pim@rhino:~/src$ sudo apt install libmnl-dev
-pim@rhino:~/src$ git clone https://github.com/pimvanpelt/lcpng.git
+pim@rhino:~/src$ git clone https://git.ipng.ch/ipng/lcpng.git
 pim@rhino:~/src$ git clone https://gerrit.fd.io/r/vpp
 pim@rhino:~/src$ ln -s ~/src/lcpng ~/src/vpp/src/plugins/lcpng
 pim@rhino:~/src$ cd ~/src/vpp

content/articles/2022-03-27-vppcfg-1.md

@@ -33,7 +33,7 @@ In this first post, let's take a look at tablestakes: writing a YAML specificati
 configuration elements of VPP, and then ensures that the YAML file is both syntactically as well as
 semantically correct.
 
-**Note**: Code is on [my Github](https://github.com/pimvanpelt/vppcfg), but it's not quite ready for
+**Note**: Code is on [my Github](https://git.ipng.ch/ipng/vppcfg), but it's not quite ready for
 prime-time yet. Take a look, and engage with us on GitHub (pull requests preferred over issues themselves)
 or reach out by [contacting us](/s/contact/).
 
@@ -348,7 +348,7 @@ to mess up my (or your!) VPP router by feeding it garbage, so the lions' share o
 has been to assert the YAML file is both syntactically and semantically valid.
 
 
-In the mean time, you can take a look at my code on [GitHub](https://github.com/pimvanpelt/vppcfg), but to
+In the mean time, you can take a look at my code on [GitHub](https://git.ipng.ch/ipng/vppcfg), but to
 whet your appetite, here's a hefty configuration that demonstrates all implemented types:
 
 ```

content/articles/2022-04-02-vppcfg-2.md

@@ -32,7 +32,7 @@ the configuration to the dataplane. Welcome to `vppcfg`!
 In this second post of the series, I want to talk a little bit about how planning a path from a running
 configuration to a desired new configuration might look like.
 
-**Note**: Code is on [my Github](https://github.com/pimvanpelt/vppcfg), but it's not quite ready for
+**Note**: Code is on [my Github](https://git.ipng.ch/ipng/vppcfg), but it's not quite ready for
 prime-time yet. Take a look, and engage with us on GitHub (pull requests preferred over issues themselves)
 or reach out by [contacting us](/s/contact/).
 

content/articles/2022-10-14-lab-1.md

@@ -275,7 +275,6 @@ that will point at an `unbound` running on `lab.ipng.ch` itself.
 I can now create any file I'd like which may use variable substition and other jinja2 style templating. Take
 for example these two files:
 
-{% raw %}
 ```
 pim@lab:~/src/lab$ cat overlays/bird/common/etc/netplan/01-netcfg.yaml.j2
 network:
@@ -292,13 +291,12 @@ network:
 
 pim@lab:~/src/lab$ cat overlays/bird/common/etc/netns/dataplane/resolv.conf.j2
 domain lab.ipng.ch
-search{% for domain in lab.nameserver.search %} {{domain}}{%endfor %}
+search{% for domain in lab.nameserver.search %} {{ domain }}{% endfor %}
 
 {% for resolver in lab.nameserver.addresses %}
-nameserver {{resolver}}
-{%endfor%}
+nameserver {{ resolver }}
+{% endfor %}
 ```
-{% endraw %}
 
 The first file is a [[NetPlan.io](https://netplan.io/)] configuration that substitutes the correct management
 IPv4 and IPv6 addresses and gateways. The second one enumerates a set of search domains and nameservers, so that

content/articles/2023-02-12-fitlet2.md

@@ -171,12 +171,12 @@ GigabitEthernet1/0/0               1     up   GigabitEthernet1/0/0
 
 After this exploratory exercise, I have learned enough about the hardware to be able to take the
 Fitlet2 out for a spin.  To configure the VPP instance, I turn to
-[[vppcfg](https://github.com/pimvanpelt/vppcfg)], which can take a YAML configuration file
+[[vppcfg](https://git.ipng.ch/ipng/vppcfg)], which can take a YAML configuration file
 describing the desired VPP configuration, and apply it safely to the running dataplane using the VPP
 API. I've written a few more posts on how it does that, notably on its [[syntax]({{< ref "2022-03-27-vppcfg-1" >}})]
 and its [[planner]({{< ref "2022-04-02-vppcfg-2" >}})].  A complete
 configuration guide on vppcfg can be found
-[[here](https://github.com/pimvanpelt/vppcfg/blob/main/docs/config-guide.md)].
+[[here](https://git.ipng.ch/ipng/vppcfg/blob/main/docs/config-guide.md)].
 
 ```
 pim@fitlet:~$ sudo dpkg -i {lib,}vpp*23.06*deb

content/articles/2023-02-24-coloclue-vpp-2.md

@@ -185,7 +185,7 @@ forgetful chipmunk-sized brain!), so here, I'll only recap what's already writte
 
 **1. BUILD:** For the first step, the build is straight forward, and yields a VPP instance based on
 `vpp-ext-deps_23.06-1` at version `23.06-rc0~71-g182d2b466`, which contains my
-[[LCPng](https://github.com/pimvanpelt/lcpng.git)] plugin. I then copy the packages to the router.
+[[LCPng](https://git.ipng.ch/ipng/lcpng.git)] plugin. I then copy the packages to the router.
 The router has an E-2286G CPU @ 4.00GHz with 6 cores and 6 hyperthreads. There's a really handy tool
 called `likwid-topology` that can show how the L1, L2 and L3 cache lines up with respect to CPU
 cores. Here I learn that CPU (0+6) and (1+7) share L1 and L2 cache -- so I can conclude that 0-5 are
@@ -351,7 +351,7 @@ in `vppcfg`:
 *   When I create the initial `--novpp` config, there's a bug in `vppcfg` where I incorrectly
     reference a dataplane object which I haven't initialized (because with `--novpp` the tool
     will not contact the dataplane at all. That one was easy to fix, which I did in [[this
-    commit](https://github.com/pimvanpelt/vppcfg/commit/0a0413927a0be6ed3a292a8c336deab8b86f5eee)]).
+    commit](https://git.ipng.ch/ipng/vppcfg/commit/0a0413927a0be6ed3a292a8c336deab8b86f5eee)]).
 
 After that small detour, I can now proceed to configure the dataplane by offering the resulting
 VPP commands, like so:
@@ -573,7 +573,7 @@ see is that which is destined to the controlplane (eg, to one of the IPv4 or IPv
 multicast/broadcast groups that they are participating in), so things like tcpdump or SNMP won't
 really work.
 
-However, due to my [[vpp-snmp-agent](https://github.com/pimvanpelt/vpp-snmp-agent.git)], which is
+However, due to my [[vpp-snmp-agent](https://git.ipng.ch/ipng/vpp-snmp-agent.git)], which is
 feeding as an AgentX behind an snmpd that in turn is running in the `dataplane` namespace, SNMP scrapes
 work as they did before, albeit with a few different interface names.
 

content/articles/2023-04-09-vpp-stats.md

@@ -14,7 +14,7 @@ performance and versatility. For those of us who have used Cisco IOS/XR devices,
 _ASR_ (aggregation service router), VPP will look and feel quite familiar as many of the approaches
 are shared between the two.
 
-I've been working on the Linux Control Plane [[ref](https://github.com/pimvanpelt/lcpng)], which you
+I've been working on the Linux Control Plane [[ref](https://git.ipng.ch/ipng/lcpng)], which you
 can read all about in my series on VPP back in 2021:
 
 [![DENOG14](/assets/vpp-stats/denog14-thumbnail.png){: style="width:300px; float: right; margin-left: 1em;"}](https://video.ipng.ch/w/erc9sAofrSZ22qjPwmv6H4)
@@ -70,7 +70,7 @@ answered by a Response PDU.
 
 Using parts of a Python Agentx library written by GitHub user hosthvo
 [[ref](https://github.com/hosthvo/pyagentx)], I tried my hands at writing one of these AgentX's.
-The resulting source code is on [[GitHub](https://github.com/pimvanpelt/vpp-snmp-agent)]. That's the
+The resulting source code is on [[GitHub](https://git.ipng.ch/ipng/vpp-snmp-agent)]. That's the
 one that's running in production ever since I started running VPP routers at IPng Networks AS8298.
 After the _AgentX_ exposes the dataplane interfaces and their statistics into _SNMP_, an open source
 monitoring tool such as LibreNMS [[ref](https://librenms.org/)] can discover the routers and draw
@@ -126,7 +126,7 @@ for any interface created in the dataplane.
 
 I wish I were good at Go, but I never really took to the language. I'm pretty good at Python, but
 sorting through the stats segment isn't super quick as I've already noticed in the Python3 based
-[[VPP SNMP Agent](https://github.com/pimvanpelt/vpp-snmp-agent)]. I'm probably the world's least
+[[VPP SNMP Agent](https://git.ipng.ch/ipng/vpp-snmp-agent)]. I'm probably the world's least
 terrible C programmer, so maybe I can take a look at the VPP Stats Client and make sense of it. Luckily,
 there's an example already in `src/vpp/app/vpp_get_stats.c` and it reveals the following pattern:
 

content/articles/2023-05-07-vpp-mpls-1.md

@@ -19,7 +19,7 @@ same time keep an IPng Site Local network with IPv4 and IPv6 that is separate fr
 based on hardware/silicon based forwarding at line rate and high availability. You can read all
 about my Centec MPLS shenanigans in [[this article]({{< ref "2023-03-11-mpls-core" >}})].
 
-Ever since the release of the Linux Control Plane [[ref](https://github.com/pimvanpelt/lcpng)]
+Ever since the release of the Linux Control Plane [[ref](https://git.ipng.ch/ipng/lcpng)]
 plugin in VPP, folks have asked "What about MPLS?" -- I have never really felt the need to go this
 rabbit hole, because I figured that in this day and age, higher level IP protocols that do tunneling
 are just as performant, and a little bit less of an 'art' to get right. For example, the Centec

content/articles/2023-05-21-vpp-mpls-3.md

@@ -459,6 +459,6 @@ and VPP, and the overall implementation before attempting to use in production.
 we got at least some of this right, but testing and runtime experience will tell.
 
 I will be silently porting the change into my own copy of the Linux Controlplane called lcpng on
-[[GitHub](https://github.com/pimvanpelt/lcpng.git)]. If you'd like to test this - reach out to the VPP
+[[GitHub](https://git.ipng.ch/ipng/lcpng.git)]. If you'd like to test this - reach out to the VPP
 Developer [[mailinglist](mailto:vpp-dev@lists.fd.io)] any time!
 

content/articles/2023-05-28-vpp-mpls-4.md

@@ -385,5 +385,5 @@ and VPP, and the overall implementation before attempting to use in production.
 we got at least some of this right, but testing and runtime experience will tell.
 
 I will be silently porting the change into my own copy of the Linux Controlplane called lcpng on
-[[GitHub](https://github.com/pimvanpelt/lcpng.git)]. If you'd like to test this - reach out to the VPP
+[[GitHub](https://git.ipng.ch/ipng/lcpng.git)]. If you'd like to test this - reach out to the VPP
 Developer [[mailinglist](mailto:vpp-dev@lists.fd.io)] any time!

content/articles/2023-10-21-vpp-ixp-gateway-1.md

@@ -304,7 +304,7 @@ Gateway, just to show a few of the more advanced features of VPP. For me, this t
 line of thinking: classifiers. This extract/match/act pattern can be used in policers, ACLs and
 arbitrary traffic redirection through VPP's directed graph (eg. selecting a next node for
 processing). I'm going to deep-dive into this classifier behavior in an upcoming article, and see
-how I might add this to [[vppcfg](https://github.com/pimvanpelt/vppcfg.git)], because I think it
+how I might add this to [[vppcfg](https://git.ipng.ch/ipng/vppcfg.git)], because I think it
 would be super powerful to abstract away the rather complex underlying API into something a little
 bit more ... user friendly. Stay tuned! :)
 

content/articles/2024-03-06-vpp-babel-1.md

@@ -359,7 +359,7 @@ does not have an IPv4 address. Except -- I'm bending the rules a little bit by d
 There's an internal function `ip4_sw_interface_enable_disable()` which is called to enable IPv4
 processing on an interface once the first IPv4 address is added. So my first fix is to force this to
 be enabled for any interface that is exposed via Linux Control Plane, notably in `lcp_itf_pair_create()`
-[[here](https://github.com/pimvanpelt/lcpng/blob/main/lcpng_interface.c#L777)].
+[[here](https://git.ipng.ch/ipng/lcpng/blob/main/lcpng_interface.c#L777)].
 
 This approach is partially effective:
 
@@ -500,7 +500,7 @@ which is unnumbered. Because I don't know for sure if everybody would find this
 I make sure to guard the behavior behind a backwards compatible configuration option.
 
 If you're curious, please take a look at the change in my [[GitHub
-repo](https://github.com/pimvanpelt/lcpng/commit/a960d64a87849d312b32d9432ffb722672c14878)], in
+repo](https://git.ipng.ch/ipng/lcpng/commit/a960d64a87849d312b32d9432ffb722672c14878)], in
 which I:
 1.   add a new configuration option, `lcp-sync-unnumbered`, which defaults to `on`. That would be
      what the plugin would do in the normal case: copy forward these borrowed IP addresses to Linux.

content/articles/2024-04-06-vpp-ospf.md

@@ -147,7 +147,7 @@ With all of that, I am ready to demonstrate two working solutions now. I first c
 Ondrej's [[commit](https://gitlab.nic.cz/labs/bird/-/commit/280daed57d061eb1ebc89013637c683fe23465e8)].
 Then, I compile VPP with my pending [[gerrit](https://gerrit.fd.io/r/c/vpp/+/40482)]. Finally,
 to demonstrate how `update_loopback_addr()` might work, I compile `lcpng` with my previous
-[[commit](https://github.com/pimvanpelt/lcpng/commit/a960d64a87849d312b32d9432ffb722672c14878)],
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/a960d64a87849d312b32d9432ffb722672c14878)],
 which allows me to inhibit copying forward addresses from VPP to Linux, when using _unnumbered_
 interfaces.
 

content/articles/2024-04-27-freeix-1.md

@@ -1,8 +1,9 @@
 ---
 date: "2024-04-27T10:52:11Z"
-title: FreeIX - Remote
+title: "FreeIX Remote - Part 1"
 aliases:
 - /s/articles/2024/04/27/freeix-1.html
+- /s/articles/2024/04/27/freeix-remote/
 ---
 
 # Introduction
@@ -91,7 +92,7 @@ their traffic to these remote internet exchanges.
 There are two types of BGP neighbor adjacency:
 
 1.   ***Members***: these are {ip-address,AS}-tuples which FreeIX has explicitly configured. Learned prefixes are added
-     to as-set AS50869:AS-MEMBERS. Members receive _all_ prefixes from FreeIX, each annotated with BGP **informational**
+     to as-set AS50869:AS-MEMBERS. Members receive _some or all_ prefixes from FreeIX, each annotated with BGP **informational**
      communities, and members can drive certain behavior with BGP **action** communities.
 
 1.   ***Peers***: these are all other entities with whom FreeIX has an adjacency at public internet exchanges or private
@@ -195,12 +196,12 @@ network interconnects:
 *   `(50869,3020,1)`: Inhibit Action (30XX), Country (3020), Switzerland (1)
 *   `(50869,3030,1308)`: Inhibit Action (30XX), IXP (3030), PeeringDB IXP for LS-IX (1308)
 
-Further actions can be placed on a per-remote-neighbor basis:
+Four actions can be placed on a per-remote-asn basis:
 
 *   `(50869,3040,13030)`: Inhibit Action (30XX), AS (3040), Init7 (AS13030)
-*   `(50869,3041,6939)`: Prepend Action (30XX), Prepend Once (3041), Hurricane Electric (AS6939)
-*   `(50869,3042,12859)`: Prepend Action (30XX), Prepend Twice (3042), BIT BV (AS12859)
-*   `(50869,3043,8283)`: Prepend Action (30XX), Prepend Three Times (3043), Coloclue (AS8283)
+*   `(50869,3100,6939)`: Prepend Once Action (3100), Hurricane Electric (AS6939)
+*   `(50869,3200,12859)`: Prepend Twice Action (3200), BIT BV (AS12859)
+*   `(50869,3300,8283)`: Prepend Thice Action (3300), Coloclue (AS8283)
 
 Peers cannot set these actions, as all action communities will be stripped on ingress.  Members can set these action
 communities on their sessions with FreeIX routers, however in some cases they may also be set by FreeIX operators when

content/articles/2024-05-25-nat64-1.md

@@ -101,6 +101,7 @@ IPv6 network and access the internet via a shared IPv6 address.
 I will assign a pool of four public IPv4 addresses and eight IPv6 addresses to each border gateway:
 
 | **Machine** | **IPv4 pool** | **IPv6 pool** |
+| ----------- | ------------- | ------------- |
 | border0.chbtl0.net.ipng.ch | <span style='color:green;'>194.126.235.0/30</span> | <span style='color:blue;'>2001:678:d78::3:0:0/125</span> |
 | border0.chrma0.net.ipng.ch | <span style='color:green;'>194.126.235.4/30</span> | <span style='color:blue;'>2001:678:d78::3:1:0/125</span> |
 | border0.chplo0.net.ipng.ch | <span style='color:green;'>194.126.235.8/30</span> | <span style='color:blue;'>2001:678:d78::3:2:0/125</span> |

content/articles/2024-06-22-vpp-ospf-2.md

@@ -250,10 +250,10 @@ remove the IPv4 and IPv6 addresses from the <span style='color:red;font-weight:b
 routers in Br&uuml;ttisellen. They are directly connected, and if anything goes wrong, I can walk
 over and rescue them. Sounds like a safe way to start!
 
-I quickly add the ability for [[vppcfg](https://github.com/pimvanpelt/vppcfg)] to configure
+I quickly add the ability for [[vppcfg](https://git.ipng.ch/ipng/vppcfg)] to configure
 _unnumbered_ interfaces. In VPP, these are interfaces that don't have an IPv4 or IPv6 address of
 their own, but they borrow one from another interface. If you're curious, you can take a look at the
-[[User Guide](https://github.com/pimvanpelt/vppcfg/blob/main/docs/config-guide.md#interfaces)] on
+[[User Guide](https://git.ipng.ch/ipng/vppcfg/blob/main/docs/config-guide.md#interfaces)] on
 GitHub.
 
 Looking at their `vppcfg` files, the change is actually very easy, taking as an example the
@@ -291,7 +291,7 @@ interface.
 
 In the article, you'll see that discussed as _Solution 2_, and it includes a bit of rationale why I
 find this better. I implemented it in this
-[[commit](https://github.com/pimvanpelt/lcpng/commit/a960d64a87849d312b32d9432ffb722672c14878)], in
+[[commit](https://git.ipng.ch/ipng/lcpng/commit/a960d64a87849d312b32d9432ffb722672c14878)], in
 case you're curious, and the commandline keyword is `lcp lcp-sync-unnumbered off` (the default is
 _on_).
 

content/articles/2024-09-08-sflow-1.md

@@ -0,0 +1,725 @@
+---
+date: "2024-09-08T12:51:23Z"
+title: 'VPP with sFlow - Part 1'
+---
+
+# Introduction
+
+{{< image float="right" src="/assets/sflow/sflow.gif" alt="sFlow Logo" width='12em' >}}
+
+In January of 2023, an uncomfortably long time ago at this point, an acquaintance of mine called
+Ciprian reached out to me after seeing my [[DENOG
+#14](https://video.ipng.ch/w/erc9sAofrSZ22qjPwmv6H4)] presentation. He was interested to learn about
+IPFIX and was asking if sFlow would be an option.  At the time, there was a plugin in VPP called
+[[flowprobe](https://s3-docs.fd.io/vpp/24.10/cli-reference/clis/clicmd_src_plugins_flowprobe.html)]
+which is able to emit IPFIX records.  Unfotunately I never really got it to work well in my tests,
+as either the records were corrupted, sub-interfaces didn't work, or the plugin would just crash the
+dataplane entirely. In the meantime, the folks at [[Netgate](https://netgate.com/)] submitted quite
+a few fixes to flowprobe, but it remains an expensive operation computationally. Wouldn't copying
+one in a thousand or ten thousand packet headers with flow _sampling_ not be just as good?
+
+In the months that followed, I discussed the feature with the incredible folks at
+[[inMon](https://inmon.com/)], the original designers and maintainers of the sFlow protocol and
+toolkit. Neil from inMon wrote a prototype and put it on [[GitHub](https://github.com/sflow/vpp)]
+but for lack of time I didn't manage to get it to work, which was largely my fault by the way.
+
+However, I have a bit of time on my hands in September and October, and just a few weeks ago,
+my buddy Pavel from [[FastNetMon](https://fastnetmon.com/)] pinged that very dormant thread about
+sFlow being a potentially useful tool for anti DDoS protection using VPP. And I very much agree!
+
+## sFlow: Protocol
+
+Maintenance of the protocol is performed by the [[sFlow.org](https://sflow.org/)] consortium, the
+authoritative source of the sFlow protocol specifications. The current version of sFlow is v5.
+
+sFlow, short for _sampled Flow_, works at the ethernet layer of the stack, where it inspects one in
+N datagrams (typically 1:1000 or 1:10000) going through the physical network interfaces of a device.
+On the device, an **sFlow Agent** does the sampling. For each sample the Agent takes, the first M
+bytes (typically 128) are copied into an sFlow Datagram. Sampling metadata is added, such as
+the ingress (or egress) interface and sampling process parameters. The Agent can then optionally add
+forwarding information (such as router source- and destination prefix, MPLS LSP information, BGP
+communties, and what-not). Finally the Agent will periodically read the octet and packet counters of
+physical network interface(s). Ultimately, the Agent will send the samples and additional
+information over the network as a UDP datagram, to an **sFlow Collector** for further processing.
+
+sFlow has been specifically designed to take advantages of the statistical properties of packet
+sampling and can be modeled using statistical sampling theory. This means that the sFlow traffic
+monitoring system will always produce statistically quantifiable measurements. You can read more
+about it in Peter Phaal and Sonia Panchen's
+[[paper](https://sflow.org/packetSamplingBasics/index.htm)], I certainly did and my head spun a
+little bit at the math :)
+
+### sFlow: Netlink PSAMPLE
+
+sFlow is meant to be a very _lightweight_ operation for the sampling equipment. It can typically be
+done in hardware, but there also exist several software implementations. One very clever thing, I
+think, is decoupling the sampler from the rest of the Agent. The Linux kernel has a packet sampling
+API called [[PSAMPLE](https://github.com/torvalds/linux/blob/master/net/psample/psample.c)], which
+allows _producers_ to send samples to a certain _group_, and then allows _consumers_ to subscribe to
+samples of a certrain _group_. The PSAMPLE API uses
+[[NetLink](https://docs.kernel.org/userspace-api/netlink/intro.html)] under the covers. The cool
+thing, for me anyway, is that I have a little bit of experience with Netlink due to my work on VPP's
+[[Linux Control Plane]({{< ref 2021-08-25-vpp-4 >}})] plugin.
+
+The idea here is that some **sFlow Agent**, notably a VPP plugin, will be taking periodic samples
+from the physical network interfaces, and producing Netlink messages. Then, some other program,
+notably outside of VPP, can consume these messages and further handle them, creating UDP packets
+with sFlow samples and counters and other information, and sending them to an **sFlow Collector**
+somewhere else on the network.
+
+{{< image width="100px" float="left" src="/assets/shared/brain.png" alt="Warning" >}}
+
+There's a handy utility called [[psampletest](https://github.com/sflow/psampletest)] which can
+subscribe to these PSAMPLE netlink groups and retrieve the samples. The first time I used all of
+this stuff, I wasn't aware of this utility and I kept on getting errors. It turns out, there's a
+kernel module that needs to be loaded: `modprobe psample` and `psampletest` helpfully does that for
+you [[ref](https://github.com/sflow/psampletest/blob/main/psampletest.c#L799)], so just make sure
+the module is loaded and added to `/etc/modules` before you spend as many hours as I did pulling out
+hair.
+
+## VPP: sFlow Plugin
+
+For the purposes of my initial testing, I'll simply take a look at Neil's prototype on
+[[GitHub](https://github.com/sflow/vpp)] and see what I learn in terms of functionality and
+performance.
+
+### sFlow Plugin: Anatomy
+
+The design is purposefully minimal, to do all of the heavy lifting outside of the VPP dataplane. The
+plugin will create a new VPP _graph node_ called `sflow`, which the operator can insert after
+`device-input`, in other words, if enabled, the plugin will get a copy of all packets that are read
+from an input provider, such as `dpdk-input` or `rdma-input`. The plugin's job is to process the
+packet, and if it's not selected for sampling, just move it onwards to the next node, typically
+`ethernet-input`. Almost all of the interesting action is in `node.c`
+
+The kicker is, that one in N packets will be selected to sample, after which:
+1.   the ethernet header (`*en`) is extracted from the packet
+1.   the input interface (`hw_if_index`) is extracted from the VPP buffer. Remember, sFlow works
+with physical network interfaces!
+1.   if there are too many samples from this worker thread being worked on, it is discarded and an
+     error counter is incremented. This protects the main thread from being slammed with samples if
+     there are simply too many being fished out of the dataplane.
+1.   Otherwise:
+     *   a new `sflow_sample_t` is created, with all the sampling process metadata filled in
+     *   the first 128 bytes of the packet are copied into the sample
+     *   an RPC is dispatched to the main thread, which will send the sample to the PSAMPLE channel
+
+Both a debug CLI command and API call are added:
+
+```
+sflow enable-disable <interface-name> [<sampling_N>]|[disable]
+```
+
+Some observations:
+
+First off, the sampling_N in Neil's demo is a global rather than per-interface setting. It would
+make sense to make this be per-interface, as routers typically have a mixture of 1G/10G and faster
+100G network cards available. It was a surprise when I set one interface to 1:1000 and the other to
+1:10000 and then saw the first interface change its sampling rate also. It's a small thing, and 
+will not be an issue to change.
+
+{{< image width="5em" float="left" src="/assets/shared/warning.png" alt="Warning" >}}
+
+Secondly, sending the RPC to main uses `vl_api_rpc_call_main_thread()`, which
+requires a _spinlock_ in `src/vlibmemory/memclnt_api.c:649`. I'm somewhat worried that when many
+samples are sent from many threads, there will be lock contention and performance will suffer.
+
+### sFlow Plugin: Functional
+
+I boot up the [[IPng Lab]({{< ref 2022-10-14-lab-1 >}})] and install a bunch of sFlow tools on it,
+make sure the `psample` kernel module is loaded.  In this first test I'll take a look at
+tablestakes. I compile VPP with the sFlow plugin, and enable that plugin in `startup.conf` on each
+of the four VPP routers.  For reference, the Lab looks like this:
+
+{{< image src="/assets/vpp-mpls/LAB v2.svg" alt="Lab Setup" >}}
+
+What I'll do is start an `iperf3` server on `vpp0-3` and then hit it from `vpp0-0`, to generate
+a few TCP traffic streams back and forth, which will be traversing `vpp0-2` and `vpp0-1`, like so:
+
+```
+pim@vpp0-3:~ $ iperf3 -s -D
+pim@vpp0-0:~ $ iperf3 -c vpp0-3.lab.ipng.ch -t 86400 -P 10 -b 10M
+```
+
+### Configuring VPP for sFlow
+
+While this `iperf3` is running, I'll log on to `vpp0-2` to take a closer look. The first thing I do,
+is turn on packet sampling on `vpp0-2`'s interface that points at `vpp0-3`, which is `Gi10/0/1`, and
+the interface that points at `vpp0-0`, which is `Gi10/0/0`. That's easy enough, and I will use a
+sampling rate of 1:1000 as these interfaces are GigabitEthernet:
+
+```
+root@vpp0-2:~# vppctl sflow enable-disable GigabitEthernet10/0/0 1000
+root@vpp0-2:~# vppctl sflow enable-disable GigabitEthernet10/0/1 1000
+root@vpp0-2:~# vppctl show run | egrep '(Name|sflow)'
+       Name         State       Calls       Vectors  Suspends      Clocks     Vectors/Call  
+sflow              active        5656         24168         0      9.01e2     4.27
+```
+
+Nice! VPP inserted the `sflow` node between `dpdk-input` and `ethernet-input` where it can do its
+business. But is it sending data? To answer this question, I can first take a look at the
+`psampletest` tool:
+
+```
+root@vpp0-2:~# psampletest 
+pstest: modprobe psample returned 0
+pstest: netlink socket number = 1637
+pstest: getFamily
+pstest: generic netlink CMD = 1
+pstest: generic family name: psample
+pstest: generic family id: 32
+pstest: psample attr type: 4 (nested=0) len: 8
+pstest: psample attr type: 5 (nested=0) len: 8
+pstest: psample attr type: 6 (nested=0) len: 24
+pstest: psample multicast group id: 9
+pstest: psample multicast group: config
+pstest: psample multicast group id: 10
+pstest: psample multicast group: packets
+pstest: psample found group packets=10
+pstest: joinGroup 10
+pstest: received Netlink ACK
+pstest: joinGroup 10
+pstest: set headers...
+pstest: serialize...
+pstest: print before sending...
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=7 out=9 n=1000 seq=1 pktlen=1514 hdrlen=31 pkt=0x558c08ba4958 q=3 depth=33333333 delay=123456
+pstest: send...
+pstest: send_psample getuid=0 geteuid=0
+pstest: sendmsg returned 140
+pstest: free...
+pstest: start read loop...
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=1 out=0 n=1000 seq=600320 pktlen=2048 hdrlen=132 pkt=0x7ffe0e4776c8 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=1 out=0 n=1000 seq=600321 pktlen=2048 hdrlen=132 pkt=0x7ffe0e4776c8 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=1 out=0 n=1000 seq=600322 pktlen=2048 hdrlen=132 pkt=0x7ffe0e4776c8 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=2 out=0 n=1000 seq=600423 pktlen=66 hdrlen=70 pkt=0x7ffdb0d5a1e8 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=1 out=0 n=1000 seq=600324 pktlen=2048 hdrlen=132 pkt=0x7ffe0e4776c8 q=0 depth=0 delay=0
+```
+
+I am amazed! The `psampletest` output shows a few packets, considering I'm asking `iperf3` to push
+100Mbit using 9000 byte jumboframes (which would be something like 1400 packets/second), I can
+expect two or three samples per second. I immediately notice a few things:
+
+***1. Network Namespace***: The Netlink sampling channel belongs to a network _namespace_. The VPP
+process is running in the _default_ netns, so its PSAMPLE netlink messages will be in that namespace. 
+Thus, the `psampletest` and other tools must also run in that namespace. I mention this because in
+Linux CP, often times the controlplane interfaces are created in a dedicated `dataplane` network
+namespace.
+
+***2. pktlen and hdrlen***: The pktlen is wrong, and this is a bug. In VPP, packets are put into
+buffers of size 2048, and if there is a jumboframe, that means multiple buffers are concatenated for
+the same packet. The packet length here ought to be 9000 in one direction. Looking at the `in=2`
+packet with length 66, that looks like a legitimate ACK packet on the way back. But why is the
+hdrlen set to 70 there? I'm going to want to ask Neil about that.
+
+***3. ingress and egress***: The `in=1` and one packet with `in=2` represent the input `hw_if_index`
+which is the ifIndex that VPP assigns to its devices. And looking at `show interfaces`, indeed
+number 1 corresponds with `GigabitEthernet10/0/0` and 2 is `GigabitEthernet10/0/1`, which checks
+out:
+```
+root@vpp0-2:~# vppctl show int
+              Name          Idx    State  MTU (L3/IP4/IP6/MPLS)     Counter          Count     
+GigabitEthernet10/0/0         1      up          9000/0/0/0     rx packets             469552764
+                                                                rx bytes           4218754400233
+                                                                tx packets             133717230
+                                                                tx bytes              8887341013
+                                                                drops                       6050
+                                                                ip4                    469321635
+                                                                ip6                       225164
+GigabitEthernet10/0/1         2      up          9000/0/0/0     rx packets             133527636
+                                                                rx bytes              8816920909
+                                                                tx packets             469353481
+                                                                tx bytes           4218736200819
+                                                                drops                       6060
+                                                                ip4                    133489925
+                                                                ip6                        29139
+
+```
+
+***4. ifIndexes are orthogonal***: These `in=1` or `in=2` ifIndex numbers are constructs of the VPP
+dataplane.  Notably, VPP's numbering of interface index is strictly _orthogonal_ to Linux, and it's
+not guaranteed that there even _exists_ an interface in Linux for the PHY upon which the sampling is
+happening. Said differently, `in=1` here is meant to reference VPP's `GigabitEthernet10/0/0`
+interface, but in Linux, `ifIndex=1` is a completely different interface (`lo`) in the default
+network namespace. Similarly `in=2` for VPP's `Gi10/0/1` interface corresponds to interface `enp1s0`
+in Linux:
+
+```
+root@vpp0-2:~# ip link
+1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
+    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
+2: enp1s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc fq_codel state UP mode DEFAULT group default qlen 1000
+    link/ether 52:54:00:f0:01:20 brd ff:ff:ff:ff:ff:ff
+```
+
+***5. Counters***: sFlow periodically polls the interface counters for all interfaces. It will
+normally use `/proc/net/` entries for that, but there are two problems with this:
+
+1.   There may not exist a Linux representation of the interface, for example if it's only doing L2
+bridging or cross connects in the VPP dataplane, and it does not have a Linux Control Plane
+interface, or `linux-cp` is not used at all.
+
+1.   Even if it does exist and it's the "correct" ifIndex in Linux, for example if the _Linux
+Interface Pair_'s tuntap `host_vif_index` index is used, even then the statistics counters in the
+Linux representation will only count packets and octets of _punted_ packets, that is to say, the
+stuff that LinuxCP has decided need to go to the Linux kernel through the TUN/TAP device.  Important
+to note that east-west traffic that goes _through_ the dataplane, is never punted to Linux, and as
+such, the counters will be undershooting: only counting traffic _to_ the router, not _through_ the
+router.
+
+### VPP sFlow: Performance
+
+Now that I've shown that Neil's proof of concept works, I will take a better look at the performance
+of the plugin. I've made a mental note that the plugin sends RPCs from worker threads to the main
+thread to marshall the PSAMPLE messages out. I'd like to see how expensive that is, in general. So,
+I pull boot two Dell R730 machines in IPng's Lab and put them to work. The first machine will run
+Cisco's T-Rex loadtester with 8x 10Gbps ports (4x dual Intel 58299), while the second (identical)
+machine will run VPP also ith 8x 10Gbps ports (2x Intel i710-DA4).
+
+I will test a bunch of things in parallel. First off, I'll test L2 (xconnect) and L3 (IPv4 routing),
+and secondly I'll test that with and without sFlow turned on. This gives me 8 ports to configure,
+and I'll start with the L2 configuration, as follows:
+
+```
+vpp# set int state TenGigabitEthernet3/0/2 up
+vpp# set int state TenGigabitEthernet3/0/3 up
+vpp# set int state TenGigabitEthernet130/0/2 up
+vpp# set int state TenGigabitEthernet130/0/3 up
+vpp# set int l2 xconnect TenGigabitEthernet3/0/2 TenGigabitEthernet3/0/3
+vpp# set int l2 xconnect TenGigabitEthernet3/0/3 TenGigabitEthernet3/0/2
+vpp# set int l2 xconnect TenGigabitEthernet130/0/2 TenGigabitEthernet130/0/3
+vpp# set int l2 xconnect TenGigabitEthernet130/0/3 TenGigabitEthernet130/0/2
+```
+
+Then, the L3 configuration looks like this:
+
+```
+vpp# lcp create TenGigabitEthernet3/0/0 host-if xe0-0
+vpp# lcp create TenGigabitEthernet3/0/1 host-if xe0-1
+vpp# lcp create TenGigabitEthernet130/0/0 host-if xe1-0
+vpp# lcp create TenGigabitEthernet130/0/1 host-if xe1-1
+vpp# set int state TenGigabitEthernet3/0/0 up
+vpp# set int state TenGigabitEthernet3/0/1 up
+vpp# set int state TenGigabitEthernet130/0/0 up
+vpp# set int state TenGigabitEthernet130/0/1 up
+vpp# set int ip address TenGigabitEthernet3/0/0 100.64.0.1/31
+vpp# set int ip address TenGigabitEthernet3/0/1 100.64.1.1/31
+vpp# set int ip address TenGigabitEthernet130/0/0 100.64.4.1/31
+vpp# set int ip address TenGigabitEthernet130/0/1 100.64.5.1/31
+vpp# ip route add 16.0.0.0/24 via 100.64.0.0
+vpp# ip route add 48.0.0.0/24 via 100.64.1.0
+vpp# ip route add 16.0.2.0/24 via 100.64.4.0
+vpp# ip route add 48.0.2.0/24 via 100.64.5.0
+vpp# ip neighbor TenGigabitEthernet3/0/0 100.64.0.0 00:1b:21:06:00:00 static
+vpp# ip neighbor TenGigabitEthernet3/0/1 100.64.1.0 00:1b:21:06:00:01 static
+vpp# ip neighbor TenGigabitEthernet130/0/0 100.64.4.0 00:1b:21:87:00:00 static
+vpp# ip neighbor TenGigabitEthernet130/0/1 100.64.5.0 00:1b:21:87:00:01 static
+```
+
+And finally, the Cisco T-Rex configuration looks like this:
+
+```
+- version: 2
+  interfaces: [ '06:00.0', '06:00.1', '83:00.0', '83:00.1', '87:00.0', '87:00.1', '85:00.0', '85:00.1' ]
+  port_info:
+    - src_mac:  00:1b:21:06:00:00
+      dest_mac: 9c:69:b4:61:a1:dc
+    - src_mac:  00:1b:21:06:00:01
+      dest_mac: 9c:69:b4:61:a1:dd
+
+    - src_mac:  00:1b:21:83:00:00 
+      dest_mac: 00:1b:21:83:00:01
+    - src_mac:  00:1b:21:83:00:01
+      dest_mac: 00:1b:21:83:00:00
+
+    - src_mac:  00:1b:21:87:00:00
+      dest_mac: 9c:69:b4:61:75:d0
+    - src_mac:  00:1b:21:87:00:01
+      dest_mac: 9c:69:b4:61:75:d1
+
+    - src_mac:  9c:69:b4:85:00:00
+      dest_mac: 9c:69:b4:85:00:01
+    - src_mac:  9c:69:b4:85:00:01
+      dest_mac: 9c:69:b4:85:00:00
+```
+
+A little note on the use of `ip neighbor` in VPP and specific `dest_mac` in T-Rex. In L2 mode,
+because the VPP interfaces will be in promiscuous mode and simply pass through any ethernet frame
+received on interface `Te3/0/2` and copy it out on `Te3/0/3` and vice-versa, there is no need to
+tinker with MAC addresses. But in L3 mode, the NIC will only accept ethernet frames addressed to its
+MAC address, so you can see that for the first port in T-Rex, I am setting `dest_mac:
+9c:69:b4:61:a1:dc` which is the MAC address of `Te3/0/0` on VPP. And then on the way out, if VPP
+wants to send traffic back to T-Rex, I'll give it a static ARP entry with `ip neighbor .. static`.
+
+With that said, I can start a baseline loadtest like so:
+{{< image width="100%" src="/assets/sflow/trex-baseline.png" alt="Cisco T-Rex: baseline" >}}
+
+T-Rex is sending 10Gbps out on all eight interfaces (four of which are L3 routing, and four of which
+are L2 xconnecting), using packet size of 1514 bytes. This amounts of roughlu 813Kpps per port, or a
+cool 6.51Mpps in total. And I can see, in this base line configuration, the VPP router is happy to
+do the work.
+
+I then enable sFlow on the second set of four ports, using a 1:1000 sampling rate:
+
+```
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/0 1000
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/1 1000
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/2 1000
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/3 1000
+```
+
+This should yield about 3'250 or so samples per second, and things look pretty great:
+
+```
+pim@hvn6-lab:~$ vppctl show err
+   Count                  Node                              Reason               Severity 
+   5034508               sflow                     sflow packets processed         error  
+      4908               sflow                      sflow packets sampled          error  
+   5034508               sflow                     sflow packets processed         error  
+      5111               sflow                      sflow packets sampled          error  
+   5034516             l2-output                      L2 output packets            error  
+   5034516              l2-input                       L2 input packets            error  
+   5034404               sflow                     sflow packets processed         error  
+      4948               sflow                      sflow packets sampled          error  
+   5034404             l2-output                      L2 output packets            error  
+   5034404              l2-input                       L2 input packets            error  
+   5034404               sflow                     sflow packets processed         error  
+      4928               sflow                      sflow packets sampled          error  
+   5034404             l2-output                      L2 output packets            error  
+   5034404              l2-input                       L2 input packets            error  
+   5034516             l2-output                      L2 output packets            error  
+   5034516              l2-input                       L2 input packets            error  
+```
+
+I can see that the `sflow packets sampled` is roughly 0.1% of the `sflow packets processed` which
+checks out. I can also see in `psampletest` a flurry of activity, so I'm happy:
+
+```
+pim@hvn6-lab:~$ sudo psampletest 
+...
+pstest: grp=1 in=9 out=0 n=1000 seq=63388 pktlen=1510 hdrlen=132 pkt=0x7ffd9e786158 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=8 out=0 n=1000 seq=63389 pktlen=1510 hdrlen=132 pkt=0x7ffd9e786158 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=11 out=0 n=1000 seq=63390 pktlen=1510 hdrlen=132 pkt=0x7ffd9e786158 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=10 out=0 n=1000 seq=63391 pktlen=1510 hdrlen=132 pkt=0x7ffd9e786158 q=0 depth=0 delay=0
+pstest: psample netlink (type=32) CMD = 0
+pstest: grp=1 in=11 out=0 n=1000 seq=63392 pktlen=1510 hdrlen=132 pkt=0x7ffd9e786158 q=0 depth=0 delay=0
+```
+
+I confirm that all four `in` interfaced (8, 9, 10 and 11) are sending samples, and those indexes
+correctly correspond to the VPP dataplane's `sw_if_index` for `TenGig130/0/0 - 3`. Sweet! On this
+machine, each TenGig network interface has its own dedicated VPP worker thread.  Considering I
+turned on sFlow sampling on four interfaces, I should see the cost I'm paying for the feature:
+
+```
+pim@hvn6-lab:~$ vppctl show run | grep -e '(Name|sflow)'
+Name                 State         Calls       Vectors Suspends      Clocks     Vectors/Call  
+sflow               active       3908218      14350684        0      9.05e1     3.67
+sflow               active       3913266      14350680        0      1.11e2     3.67
+sflow               active       3910828      14350687        0      1.08e2     3.67
+sflow               active       3909274      14350692        0      5.66e1     3.67
+```
+
+Alright, so for the 999 packets that went through and the one packet that got sampled, on average
+VPP is spending between 90 and 111 CPU cycles per packet, and the loadtest looks squeaky clean on
+T-Rex.
+
+### VPP sFlow: Cost of passthru
+
+I decide to take a look at two edge cases. What if there are no samples being taken at all, and the
+`sflow` node is merely passing through all packets to `ethernet-input`? To simulate this, I will set
+up a bizarrely high sampling rate, say one in ten million. I'll also make the T-Rex loadtester use
+only four ports, in other words, a unidirectional loadtest, and I'll make it go much faster by
+sending smaller packets, say 128 bytes:
+
+```
+tui>start -f stl/ipng.py -p 0 2 4 6 -m 99% -t size=128
+
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/0 1000 disable
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/1 1000 disable
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/2 1000 disable
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/3 1000 disable
+
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/0 10000000
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/2 10000000
+```
+
+The loadtester is now sending 33.5Mpps or thereabouts (4x 8.37Mpps), and I can confirm that the
+`sFlow` plugin is not sampling many packets:
+
+```
+pim@hvn6-lab:~$ vppctl show err
+   Count                  Node                              Reason               Severity 
+  59777084               sflow                     sflow packets processed         error  
+         6               sflow                      sflow packets sampled          error  
+  59777152             l2-output                      L2 output packets            error  
+  59777152              l2-input                       L2 input packets            error  
+  59777104               sflow                     sflow packets processed         error  
+         6               sflow                      sflow packets sampled          error  
+  59777104             l2-output                      L2 output packets            error  
+  59777104              l2-input                       L2 input packets            error  
+
+pim@hvn6-lab:~$ vppctl show run | grep -e '(Name|sflow)'
+Name                 State         Calls       Vectors Suspends      Clocks     Vectors/Call  
+sflow                active      8186642     369674664        0      1.35e1     45.16
+sflow                active     25173660     369674696        0      1.97e1     14.68
+```
+Two observations:
+
+1.   One of these is busier than the other. Without looking further, I can already predict that the
+top one (doing 45.16 vectors/call) is the L3 thread. Reasoning: the L3 code path through the
+dataplane is a lot more expensive than 'merely' L2 XConnect. As such, the packets will spend more
+time, and therefore the iterations of the `dpdk-input` loop will be further apart in time.  And
+because of that, it'll end up consuming more packets on each subsequent iteration, in order to catch
+up.  The L2 path on the other hand, is quicker and therefore will have less packets waiting on
+subsequent iterations of `dpdk-input`.
+
+2.   The `sflow` plugin spends between 13.5 and 19.7 CPU cycles shoveling the packets into
+`ethernet-input` without doing anything to them. That's pretty low! And the L3 path is a little bit
+more efficient per packet, which is very likely because it gets to amort its L1/L2 CPU instruction
+cache over 45 packets each time it runs, while the L2 path can only amort its instruction cache over
+15 or so packets each time it runs.
+
+I let the loadtest run overnight,and the proof is in the pudding: sFlow enabled but not sampling
+works just fine:
+
+{{< image width="100%" src="/assets/sflow/trex-passthru.png" alt="Cisco T-Rex: passthru" >}}
+
+### VPP sFlow: Cost of sampling
+
+The other interesting case is to figure out how much CPU it takes to execute the code path
+with the actual sampling. This one turns out a bit trickier to measure. While leaving the previous
+loadtest running at 33.5Mpps, I disable sFlow and then re-enable it at an abnormally _high_ ratio of
+1:10 packets:
+
+```
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/0 disable
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/2 disable
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/0 10
+pim@hvn6-lab:~$ vppctl sflow enable-disable TenGigabitEthernet130/0/2 10
+```
+
+The T-Rex view immediately reveals that VPP is not doing very well, as the throughput went from
+33.5Mpps all the way down to 7.5Mpps. Ouch! Looking at the dataplane:
+
+```
+pim@hvn6-lab:~$ vppctl show err | grep sflow
+ 340502528               sflow                     sflow packets processed         error  
+  12254462               sflow                      sflow packets dropped          error  
+  22611461               sflow                      sflow packets sampled          error  
+ 422527140               sflow                     sflow packets processed         error  
+   8533855               sflow                      sflow packets dropped          error  
+  34235952               sflow                      sflow packets sampled          error  
+```
+
+Ha, this new safeguard popped up: remember all the way at the beginning, I explained how there's a
+safety net in the `sflow` plugin that will pre-emptively drop the sample if the RPC channel towards
+the main thread is seeing too many outstanding RPCs? That's happening right now, under the moniker
+`sflow packets dropped`, and it's roughly *half* of the samples.
+
+My first attempt is to back off the loadtester to roughly 1.5Mpps per port (so 6Mpps in total, under the
+current limit of 7.5Mpps), but I'm disappointed: the VPP instance is now returning 665Kpps per port
+only, which is horrible, and it's still dropping samples.
+
+My second attempt is to turn off all ports but last pair (the L2XC port), which returns 930Kpps from
+the offered 1.5Mpps. VPP is clearly not having a good time here.
+
+Finally, as a validation, I turn off all ports but the first pair (the L3 port, without sFlow), and
+ramp up the traffic to 8Mpps. Success (unsurprising to me). I also ramp up the second pair (the L2XC
+port, without sFlow), VPP forwards all 16Mpps and is happy again.
+
+Once I turn on the third pair (the L3 port, _with_ sFlow), even at 1Mpps, the whole situation
+regresses again: First two ports go down from 8Mpps to 5.2Mpps each; the third (offending) port
+delivers 740Kpps out of 1Mpps. Clearly, there's some work to do under high load situations!
+
+#### Reasoning about the bottle neck
+
+But how expensive is sending samples, really? To try to get at least some pseudo-scientific answer I
+turn off all ports again, and ramp up the one port pair with (L3 + sFlow at 1:10 ratio) to full line
+rate: that is 64 byte packets at 14.88Mpps:
+
+```
+tui>stop
+tui>start -f stl/ipng.py -m 100% -p 4 -t size=64
+```
+
+VPP is now on the struggle bus and is returning 3.16Mpps or 21% of that. But, I think it'll give me
+some reasonable data to try to feel out where the bottleneck is.
+
+```
+Thread 2 vpp_wk_1 (lcore 3)
+Time 6.3, 10 sec internal node vector rate 256.00 loops/sec 27310.73
+  vector rates in 3.1607e6, out 3.1607e6, drop 0.0000e0, punt 0.0000e0
+             Name                 State     Calls    Vectors   Suspends   Clocks   Vectors/Call  
+TenGigabitEthernet130/0/1-outp   active     77906   19943936          0   5.79e0         256.00
+TenGigabitEthernet130/0/1-tx     active     77906   19943936          0   6.88e1         256.00
+dpdk-input                       polling    77906   19943936          0   4.41e1         256.00
+ethernet-input                   active     77906   19943936          0   2.21e1         256.00
+ip4-input                        active     77906   19943936          0   2.05e1         256.00
+ip4-load-balance                 active     77906   19943936          0   1.07e1         256.00
+ip4-lookup                       active     77906   19943936          0   1.98e1         256.00
+ip4-rewrite                      active     77906   19943936          0   1.97e1         256.00
+sflow                            active     77906   19943936          0   6.14e1         256.00
+
+pim@hvn6-lab:pim# vppctl show err | grep sflow
+ 551357440               sflow                     sflow packets processed         error  
+  19829380               sflow                      sflow packets dropped          error  
+  36613544               sflow                      sflow packets sampled          error  
+```
+
+OK, the `sflow` plugin saw 551M packets, selected 36.6M of them for sampling, but ultimately only
+sent RPCs to the main thread for 16.8M samples after having dropped 19.8M of them. There are three
+code paths, each one extending the other:
+
+1.   Super cheap: pass through. I already learned that it takes about X=13.5 CPU cycles to pass
+through a packet.
+1.   Very cheap: select sample and construct the RPC, but toss it, costing Y CPU cycles.
+1.   Expensive: select sample, and send the RPC. Z CPU cycles in worker, and another amount in main.
+
+Now I don't know what Y is, but seeing as the selection only copies some data from the VPP buffer
+into a new `sflow_sample_t`, and it uses `clip_memcpy_fast()` for the sample header, I'm going to
+assume it's not _drastically_ more expensive than the super cheap case, so for simplicity I'll
+guesstimate that it takes Y=20 CPU cyces.
+
+With that guess out of the way, I can see what the `sflow` plugin is consuming for the third case:
+
+```
+AvgClocks = (Total * X + Sampled * Y + RPCSent * Z) / Total
+
+61.4 = ( 551357440 * 13.5 + 36613544 * 20 + (36613544-19829380) * Z ) / 551357440
+61.4 = ( 7443325440 + 732270880 + 16784164 * Z ) / 551357440
+33853346816 = 7443325440 + 732270880 + 16784164 * Z
+25677750496 = 16784164 * Z 
+Z = 1529
+```
+
+Good to know! I find spending O(1500) cycles to send the sample pretty reasonable. However, for a
+dataplane that is trying to do 10Mpps per core, and a core running 2.2GHz, there are really only 220
+CPU cycles to spend end-to-end. Spending an order of magnitude more than that once every ten packets
+feels dangerous to me.
+
+Here's where I start my conjecture. If I count the CPU cycles spent in the table above, I will see
+273 CPU cycles spent on average per packet. The CPU in the VPP router is an `E5-2696 v4 @ 2.20GHz`,
+which means it should be able to do `2.2e10/273 = 8.06Mpps` per thread, more than double that what I
+observe (3.16Mpps)! But, for all the `vector rates in` (3.1607e6), it also managed to emit the
+packets back out (same number: 3.1607e6).
+
+So why isn't VPP getting more packets from DPDK?  I poke around a bit and find an important clue:
+
+```
+pim@hvn6-lab:~$ vppctl show hard TenGigabitEthernet130/0/0 | grep rx\ missed; \
+                sleep 10; vppctl show hard TenGigabitEthernet130/0/0 | grep rx\ missed
+    rx missed                                     4065539464
+    rx missed                                     4182788310
+```
+
+In those ten seconds, VPP missed (4182788310-4065539464)/10 = 11.72Mpps. I already measured that it
+forwarded 3.16Mpps and you know what? 11.7 + 3.16 is precisely 14.88Mpps. All packets are accounted
+for! It's just, DPDK never managed to read them from the hardware: `sad-trombone.wav`
+
+
+As a validation, I turned off sFlow while keeping that one port at 14.88Mpps. Now, 10.8Mpps were
+delivered:
+
+```
+Thread 2 vpp_wk_1 (lcore 3)
+Time 14.7, 10 sec internal node vector rate 256.00 loops/sec 40622.64
+  vector rates in 1.0794e7, out 1.0794e7, drop 0.0000e0, punt 0.0000e0
+             Name                 State     Calls    Vectors   Suspends   Clocks   Vectors/Call  
+TenGigabitEthernet130/0/1-outp   active    620012  158723072          0   5.66e0         256.00
+TenGigabitEthernet130/0/1-tx     active    620012  158723072          0   7.01e1         256.00
+dpdk-input                       polling   620012  158723072          0   4.39e1         256.00
+ethernet-input                   active    620012  158723072          0   1.56e1         256.00
+ip4-input-no-checksum            active    620012  158723072          0   1.43e1         256.00
+ip4-load-balance                 active    620012  158723072          0   1.11e1         256.00
+ip4-lookup                       active    620012  158723072          0   2.00e1         256.00
+ip4-rewrite                      active    620012  158723072          0   2.02e1         256.00
+```
+
+Total Clocks: 201 per packet; 2.2GHz/201 = 10.9Mpps, and I am observing 10.8Mpps. As [[North of the
+Border](https://www.youtube.com/c/NorthoftheBorder)] would say: "That's not just good, it's good
+_enough_!"
+
+For completeness, I turned on all eight ports again, at line rate (8x14.88 = 119Mpps 🥰), and saw that
+about 29Mpps of that made it through. Interestingly, what was 3.16Mpps in the single-port line rate
+loadtest, went up slighty to 3.44Mpps now. What puzzles me even more, is that the non-sFlow worker
+threads are also impacted. I spent some time thinking about this and poking around, but I did not
+find a good explanation why port pair 0 (L3, no sFlow) and 1 (L2XC, no sFlow) would be impacted.
+Here's a screenshot of VPP on the struggle bus:
+
+{{< image width="100%" src="/assets/sflow/trex-overload.png" alt="Cisco T-Rex: overload at line rate" >}}
+
+**Hypothesis**: Due to the _spinlock_ in `vl_api_rpc_call_main_thread()`, the worker CPU is pegged
+for a longer time, during which the `dpdk-input` PMD can't run, so it misses out on these sweet
+sweet packets that the network card had dutifully received for it, resulting in the `rx-miss`
+situation.  While VPP's performance measurement shows 273 CPU cycles per packet and 3.16Mpps, this
+accounts only for 862M cycles, while the thread has 2200M cycles, leaving a whopping 60% of CPU
+cycles unused in the dataplane. I still don't understand why _other_ worker threads are impacted,
+though.
+
+## What's Next
+
+I'll continue to work with the folks in the sFlow and VPP communities and iterate on the plugin and
+other **sFlow Agent** machinery. In an upcoming article, I hope to share more details on how to tie
+the VPP plugin in to the `hsflowd` host sflow daemon in a way that the interface indexes, counters
+and packet lengths are all correct. Of course, the main improvement that we can make is to allow for
+the system to work better under load, which will take some thinking.
+
+I should do a few more tests with a debug binary and profiling turned on. I quickly ran a `perf`
+over the VPP (release / optimized)  binary running on the bench, but it merely said 80% of time was
+spent in `libvlib` rather than `libvnet` in the baseline (sFlow turned off).
+
+```
+root@hvn6-lab:/home/pim# perf record -p 1752441 sleep 10
+root@hvn6-lab:/home/pim# perf report --stdio --sort=dso 
+# Overhead  Shared Object (sFlow)   Overhead  Shared Object (baseline)
+# ........  ......................  ........  ........................
+#
+    79.02%  libvlib.so.24.10          54.27%  libvlib.so.24.10      
+    12.82%  libvnet.so.24.10          33.91%  libvnet.so.24.10      
+     3.77%  dpdk_plugin.so            10.87%  dpdk_plugin.so        
+     3.21%  [kernel.kallsyms]          0.81%  [kernel.kallsyms]     
+     0.29%  sflow_plugin.so            0.09%  ld-linux-x86-64.so.2  
+     0.28%  libvppinfra.so.24.10       0.03%  libc.so.6 
+     0.21%  libc.so.6                  0.01%  libvppinfra.so.24.10  
+     0.17%  libvlibapi.so.24.10        0.00%  libvlibmemory.so.24.10
+     0.15%  libvlibmemory.so.24.10     
+     0.07%  ld-linux-x86-64.so.2  
+     0.00%  vpp                   
+     0.00%  [vdso]                
+     0.00%  libsvm.so.24.10       
+```
+
+Unfortunately, I'm not much of a profiler expert, being merely a network engineer :) so I may have
+to ask for help. Of course, if you're reading this, you may also _offer_ help! There's lots of
+interesting work to do on this `sflow` plugin, with matching ifIndex for consumers like `hsflowd`,
+reading interface counters from the dataplane (or from the Prometheus Exporter), and most
+importantly, ensuring it works well, or fails gracefully, under stringent load.
+
+From the _cray-cray_ ideas department, what if we:
+1.   In worker thread, produced the sample but instead of sending an RPC to main and taking the
+lock, append it to a producer sample queue and move on. This way, no locks are needed, and each
+worker thread will have its own producer queue.
+
+1.   Create a separate worker (or even pool of workers), running on possibly a different CPU (or in
+main), that runs a loop iterating on all sflow sample queues consuming the samples and sending them
+in batches to the PSAMPLE Netlink group, possibly dropping samples if there are too many coming in.
+
+I'm reminded that this pattern exists already -- async crypto workers create a `crypto-dispatch`
+node that acts as poller for inbound crypto, and it hands off the result back into the worker
+thread: lockless at the expense of some complexity!
+
+## Acknowledgements
+
+The plugin I am testing here is a prototype written by Neil McKee of inMon. I also wanted to say
+thanks to Pavel Odintsov of FastNetMon and Ciprian Balaceanu for showing an interest in this plugin,
+and Peter Phaal for facilitating a get-together last year.
+
+Who's up for making this thing a reality?!

content/articles/2024-10-06-sflow-2.md

@@ -0,0 +1,547 @@
+---
+date: "2024-10-06T07:51:23Z"
+title: 'VPP with sFlow - Part 2'
+---
+
+# Introduction
+
+{{< image float="right" src="/assets/sflow/sflow.gif" alt="sFlow Logo" width='12em' >}}
+
+Last month, I picked up a project together with Neil McKee of [[inMon](https://inmon.com/)], the
+care takers of [[sFlow](https://sflow.org)]: an industry standard technology for monitoring high speed switched
+networks. `sFlow` gives complete visibility into the use of networks enabling performance optimization,
+accounting/billing for usage, and defense against security threats.
+
+The open source software dataplane [[VPP](https://fd.io)] is a perfect match for sampling, as it
+forwards packets at very high rates using underlying libraries like [[DPDK](https://dpdk.org/)] and
+[[RDMA](https://en.wikipedia.org/wiki/Remote_direct_memory_access)]. A clever design choice in the so
+called _Host sFlow Daemon_ [[host-sflow](https://github.com/sflow/host-sflow)], which allows for a small
+portion of code to _grab_ the samples, for example in a merchant silicon ASIC or FPGA, but also in the
+VPP software dataplane, and then _transmit_ these samples using a Linux kernel feature called
+[[PSAMPLE](https://github.com/torvalds/linux/blob/master/net/psample/psample.c)]. This greatly
+reduces the complexity of code to be implemented in the forwarding path, while at the same time
+bringing consistency to the `sFlow` delivery pipeline by (re)using the `hsflowd` business logic for
+the more complex state keeping, packet marshalling and transmission from the _Agent_ to a central
+_Collector_.
+
+Last month, Neil and I discussed the proof of concept [[ref](https://github.com/sflow/vpp-sflow/)]
+and I described this in a [[first article]({{< ref 2024-09-08-sflow-1.md >}})]. Then, we iterated on
+the VPP plugin, playing with a few different approaches to strike a balance between performance, code
+complexity, and agent features. This article describes our journey.
+
+## VPP: an sFlow plugin
+
+There are three things Neil and I specifically take a look at:
+
+1.  If `sFlow` is not enabled on a given interface, there should not be a regression on other
+interfaces.
+1.  If `sFlow` _is_ enabled, but a packet is not sampled, the overhead should be as small as
+possible, targetting single digit CPU cycles per packet in overhead.
+1.  If `sFlow` actually selects a packet for sampling, it should be moved out of the dataplane as
+quickly as possible, targetting double digit CPU cycles per sample.
+
+For all these validation and loadtests, I use a bare metal VPP machine which is receiving load from
+a T-Rex loadtester on eight TenGig ports. I have configured VPP and T-Rex as follows.
+
+**1. RX Queue Placement**
+
+It's important that the network card that is receiving the traffic, gets serviced by a worker thread
+on the same NUMA domain. Since my machine has two processors (and thus, two NUMA nodes), I will
+align the NIC with the correct processor, like so:
+
+```
+set interface rx-placement TenGigabitEthernet3/0/0 queue 0 worker 0
+set interface rx-placement TenGigabitEthernet3/0/1 queue 0 worker 2
+set interface rx-placement TenGigabitEthernet3/0/2 queue 0 worker 4
+set interface rx-placement TenGigabitEthernet3/0/3 queue 0 worker 6
+
+set interface rx-placement TenGigabitEthernet130/0/0 queue 0 worker 1
+set interface rx-placement TenGigabitEthernet130/0/1 queue 0 worker 3
+set interface rx-placement TenGigabitEthernet130/0/2 queue 0 worker 5
+set interface rx-placement TenGigabitEthernet130/0/3 queue 0 worker 7
+```
+
+**2. L3 IPv4/MPLS interfaces**
+
+I will take two pairs of interfaces, one on NUMA0, and the other on NUMA1, so that I can make a
+comparison with L3 IPv4 or MPLS running _without_ `sFlow` (these are TenGig3/0/*, which I will call
+the _baseline_ pairs) and two which are running _with_ `sFlow` (these are TenGig130/0/*, which I'll
+call the _experiment_ pairs).
+
+```
+comment { L3: IPv4 interfaces }
+set int state TenGigabitEthernet3/0/0 up
+set int state TenGigabitEthernet3/0/1 up
+set int state TenGigabitEthernet130/0/0 up
+set int state TenGigabitEthernet130/0/1 up
+set int ip address TenGigabitEthernet3/0/0 100.64.0.1/31
+set int ip address TenGigabitEthernet3/0/1 100.64.1.1/31
+set int ip address TenGigabitEthernet130/0/0 100.64.4.1/31
+set int ip address TenGigabitEthernet130/0/1 100.64.5.1/31
+ip route add 16.0.0.0/24 via 100.64.0.0
+ip route add 48.0.0.0/24 via 100.64.1.0
+ip route add 16.0.2.0/24 via 100.64.4.0
+ip route add 48.0.2.0/24 via 100.64.5.0
+ip neighbor TenGigabitEthernet3/0/0 100.64.0.0 00:1b:21:06:00:00 static
+ip neighbor TenGigabitEthernet3/0/1 100.64.1.0 00:1b:21:06:00:01 static
+ip neighbor TenGigabitEthernet130/0/0 100.64.4.0 00:1b:21:87:00:00 static
+ip neighbor TenGigabitEthernet130/0/1 100.64.5.0 00:1b:21:87:00:01 static
+```
+
+Here, the only specific trick worth mentioning is the use of `ip neighbor` to pre-populate the L2
+adjacency for the T-Rex loadtester. This way, VPP knows which MAC address to send traffic to, in
+case a packet has to be forwarded to 100.64.0.0 or 100.64.5.0. It avoids VPP from having to use ARP
+resolution.
+
+The configuration for an MPLS label switching router _LSR_ or also called _P-Router_ is added:
+
+```
+comment { MPLS interfaces }
+mpls table add 0
+set interface mpls TenGigabitEthernet3/0/0 enable
+set interface mpls TenGigabitEthernet3/0/1 enable
+set interface mpls TenGigabitEthernet130/0/0 enable
+set interface mpls TenGigabitEthernet130/0/1 enable
+mpls local-label add 16 eos via 100.64.1.0 TenGigabitEthernet3/0/1 out-labels 17
+mpls local-label add 17 eos via 100.64.0.0 TenGigabitEthernet3/0/0 out-labels 16
+mpls local-label add 20 eos via 100.64.5.0 TenGigabitEthernet130/0/1 out-labels 21
+mpls local-label add 21 eos via 100.64.4.0 TenGigabitEthernet130/0/0 out-labels 20
+```
+
+**3. L2 CrossConnect interfaces**
+
+Here, I will also use NUMA0 as my baseline (`sFlow` disabled) pair, and an equivalent pair of TenGig
+interfaces on NUMA1 as my experiment (`sFlow` enabled) pair. This way, I can both make a comparison
+on the performance impact of enabling `sFlow`, but I can also assert if any regression occurs in the
+_baseline_ pair if I enable a feature in the _experiment_ pair, which should really never happen.
+
+```
+comment { L2 xconnected interfaces }
+set int state TenGigabitEthernet3/0/2 up
+set int state TenGigabitEthernet3/0/3 up
+set int state TenGigabitEthernet130/0/2 up
+set int state TenGigabitEthernet130/0/3 up
+set int l2 xconnect TenGigabitEthernet3/0/2 TenGigabitEthernet3/0/3
+set int l2 xconnect TenGigabitEthernet3/0/3 TenGigabitEthernet3/0/2
+set int l2 xconnect TenGigabitEthernet130/0/2 TenGigabitEthernet130/0/3
+set int l2 xconnect TenGigabitEthernet130/0/3 TenGigabitEthernet130/0/2
+```
+
+**4. T-Rex Configuration**
+
+The Cisco T-Rex loadtester is running on another machine in the same rack. Physically, it has eight
+ports which are connected to a LAB switch, a cool Mellanox SN2700 running Debian [[ref]({{< ref
+2023-11-11-mellanox-sn2700.md >}})]. From there, eight ports go to my VPP machine. The LAB switch
+just has VLANs with two ports in each: VLAN 100 takes T-Rex port0 and connects it to TenGig3/0/0,
+VLAN 101 takes port1 and connects it to TenGig3/0/1, and so on. In total, sixteen ports and eight
+VLANs are used.
+
+The configuration for T-Rex then becomes:
+
+```
+- version: 2
+  interfaces: [ '06:00.0', '06:00.1', '83:00.0', '83:00.1', '87:00.0', '87:00.1', '85:00.0', '85:00.1' ]
+  port_info:
+    - src_mac:  00:1b:21:06:00:00
+      dest_mac: 9c:69:b4:61:a1:dc
+    - src_mac:  00:1b:21:06:00:01
+      dest_mac: 9c:69:b4:61:a1:dd
+
+    - src_mac:  00:1b:21:83:00:00
+      dest_mac: 00:1b:21:83:00:01
+    - src_mac:  00:1b:21:83:00:01
+      dest_mac: 00:1b:21:83:00:00
+
+    - src_mac:  00:1b:21:87:00:00
+      dest_mac: 9c:69:b4:61:75:d0
+    - src_mac:  00:1b:21:87:00:01
+      dest_mac: 9c:69:b4:61:75:d1
+
+    - src_mac:  9c:69:b4:85:00:00
+      dest_mac: 9c:69:b4:85:00:01
+    - src_mac:  9c:69:b4:85:00:01
+      dest_mac: 9c:69:b4:85:00:00
+```
+
+Do you see how the first pair sends from `src_mac` 00:1b:21:06:00:00? That's the T-Rex side, and it
+encodes the PCI device `06:00.0` in the MAC address. It sends traffic to `dest_mac`
+9c:69:b4:61:a1:dc, which is the MAC address of VPP's TenGig3/0/0 interface. Looking back at the `ip
+neighbor` VPP config above, it becomes much easier to see who is sending traffic to whom.
+
+For L2XC, the MAC addresses don't matter. VPP will set the NIC in _promiscuous_ mode which means
+it'll accept any ethernet frame, not only those sent to the NIC's own MAC address. Therefore, in
+L2XC modes (the second and fourth pair), I just use the MAC addresses from T-Rex. I find debugging
+connections and looking up FDB entries on the Mellanox switch much, much easier this way.
+
+With all config in place, but with `sFlow` disabled, I run a quick bidirectional loadtest using 256b
+packets at line rate, which shows 79.83Gbps and 36.15Mpps. All ports are forwarding, with MPLS,
+IPv4, and L2XC. Neat!
+
+{{< image src="/assets/sflow/trex-acceptance.png" alt="T-Rex Acceptance Loadtest" >}}
+
+The name of the game is now to do a loadtest that shows the packet throughput and CPU cycles spent
+for each of the plugin iterations, comparing their performance on ports with and without `sFlow`
+enabled. For each iteration, I will use exactly the same VPP configuration, I will generate
+unidirectional 4x14.88Mpps of traffic with T-Rex, and I will report on VPP's performance in
+_baseline_ and a somewhat unfavorable 1:100 sampling rate.
+
+Ready? Here I go!
+
+### v1: Workers send RPC to main
+
+***TL/DR***: _13 cycles/packet on passthrough, 4.68Mpps L2, 3.26Mpps L3, with severe regression in
+baseline_
+
+The first iteration goes all the way back to a proof of concept from last year. It's described in
+detail in my [[first post]({{< ref 2024-09-08-sflow-1.md >}})]. The performance results are not
+stellar:
+*   ☢ When slamming a single sFlow enabled interface, _all interfaces_ regress. When sending 8Mpps
+of IPv4 traffic through an _baseline_ interface, that is an interface _without_ sFlow enabled, only
+5.2Mpps get through. This is considered a mortal sin in VPP-land.
+*   ✅ Passing through packets without sampling them, costs about 13 CPU cycles, not bad.
+*   ❌ Sampling a packet, specifically at higher rates (say, 1:100 or worse, 1:10) completely
+destroys throughput. When sending 4x14.88MMpps of traffic, only one third makes it through.
+
+Here's the bloodbath as seen from T-Rex:
+
+{{< image src="/assets/sflow/trex-v1.png" alt="T-Rex Loadtest for v1" >}}
+
+**Debrief**: When we talked through these issues, we sort of drew the conclusion that it would be much
+faster if, when a worker thread produces a sample, instead of sending an RPC to main and taking the
+spinlock, that the worker appends the sample to a producer queue and moves on. This way, no locks
+are needed, and each worker thread will have its own producer queue.
+
+Then, we can create a separate thread (or even pool of threads), scheduling on possibly a different
+CPU (or in main), that runs a loop iterating on all sflow sample queues, consuming the samples and
+sending them in batches to the PSAMPLE Netlink group, possibly dropping samples if there are too
+many coming in.
+
+### v2: Workers send PSAMPLE directly
+
+**TL/DR**: _7.21Mpps IPv4 L3, 9.45Mpps L2XC, 87 cycles/packet, no impact on disabled interfaces_
+
+But before we do that, we have one curiosity itch to scratch - what if we sent the sample directly
+from the worker? With such a model, if it works, we will need no RPCs or sample queue at all. Of
+course, in this model any sample will have to be rewritten into a PSAMPLE packet and written via the
+netlink socket. It would be less complex, but not as efficient as it could be. One thing is prety
+certain, though: it should be much faster than sending an RPC to the main thread.
+
+After short refactor, Neil commits [[d278273](https://github.com/sflow/vpp-sflow/commit/d278273)],
+which adds compiler macros `SFLOW_SEND_FROM_WORKER` (v2) and `SFLOW_SEND_VIA_MAIN` (v1). When
+workers send directly, they will invoke `sflow_send_sample_from_worker()` instead of sending an RPC
+with `vl_api_rpc_call_main_thread()` in the previous version.
+
+The code currently uses `clib_warning()` to print stats from the dataplane, which is pretty
+expensive. We should be using the VPP logging framework, but for the time being, I add a few CPU
+counters so we can more accurately count the cummulative time spent for each part of the calls, see
+[[6ca61d2](https://github.com/sflow/vpp-sflow/commit/6ca61d2)]. I can now see these with `vppctl show
+err` instead.
+
+When loadtesting this, the deadly sin of impacting performance of interfaces that did not have
+`sFlow` enabled is gone. The throughput is not great, though. Instead of showing screenshots of
+T-Rex, I can also take a look at the throughput as measured by VPP itself. In its `show runtime`
+statistics, each worker thread shows both CPU cycles spent, as well as how many packets/sec it
+received and how many it transmitted:
+
+```
+pim@hvn6-lab:~$ export C="v2-100"; vppctl clear run; vppctl clear err; sleep 30; \
+                vppctl show run > $C-runtime.txt; vppctl show err > $C-err.txt
+pim@hvn6-lab:~$ grep 'vector rates' v2-100-runtime.txt  | grep -v 'in 0'
+  vector rates in 1.0909e7, out 1.0909e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 7.2078e6, out 7.2078e6, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4866e7, out 1.4866e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 9.4476e6, out 9.4476e6, drop 0.0000e0, punt 0.0000e0
+pim@hvn6-lab:~$ grep 'sflow' v2-100-runtime.txt 
+Name            State        Calls      Vectors  Suspends  Clocks    Vectors/Call  
+sflow           active      844916    216298496         0  8.69e1          256.00
+sflow           active     1107466    283511296         0  8.26e1          256.00
+pim@hvn6-lab:~$ grep -i sflow v2-100-err.txt 
+ 217929472               sflow                     sflow packets processed         error  
+   1614519               sflow                      sflow packets sampled          error  
+2606893106               sflow                    CPU cycles in sent samples       error  
+ 280697344               sflow                     sflow packets processed         error  
+   2078203               sflow                      sflow packets sampled          error  
+1844674406               sflow                    CPU cycles in sent samples       error  
+```
+
+At a glance, I can see in the first `grep`, the in and out vector (==packet) rates for each worker
+thread that is doing meaningful work (ie. has more than 0pps of input). Remember that I pinned the
+RX queues to worker threads, and this now pays dividend: worker thread 0 is servicing TenGig3/0/0
+(as _even_ worker thread numbers are on NUMA domain 0), worker thread 1 is servicing TenGig130/0/0.
+What's cool about this, is it gives me an easy way to compare baseline L3 (10.9Mpps) with experiment
+L3 (7.21Mpps). Equally, L2XC comes in at 14.88Mpps in baseline and 9.45Mpps in experiment.
+
+Looking at the output of `vppctl show error`, I can learn another interesting detail. See how there
+are 1614519 sampled packets out of 217929472 processed packets (ie. a roughly 1:100 rate)? I added a
+CPU clock cycle counter that counts cummulative clocks spent once samples are taken. I can see that
+VPP spent 2606893106 CPU cycles sending these samples. That's **1615 CPU cycles** per sent sample,
+which is pretty terrible.
+
+**Debrief**: We both understand that assembling and `send()`ing the netlink messages from within the
+dataplane is a pretty bad idea. But it's great to see that removing the use of RPCs immediately
+improves performance on non-enabled interfaces, and we learned what the cost is of sending those
+samples. An easy step forward from here is to create a producer/consumer queue, where the workers
+can just copy the packet into a queue or ring buffer, and have an external `pthread` consume from
+the queue/ring in another thread that won't block the dataplane.
+
+### v3: SVM FIFO from workers, dedicated PSAMPLE pthread
+
+**TL/DR:** _9.34Mpps L3, 13.51Mpps L2XC, 16.3 cycles/packet, but with corruption on the FIFO queue messages_
+
+Neil checks in after committing [[7a78e05](https://github.com/sflow/vpp-sflow/commit/7a78e05)]
+that he has introduced a macro `SFLOW_SEND_FIFO` which tries this new approach. There's a pretty
+elaborate FIFO queue implementation in `svm/fifo_segment.h`. Neil uses this to create a segment
+called `fifo-sflow-worker`, to which the worker can write its samples in the dataplane node. A new
+thread called `spt_process_samples` can then call `svm_fifo_dequeue()` from all workers' queues and
+pump those into Netlink.
+
+The overhead of copying the samples onto a VPP native `svm_fifo` seems to be two orders of magnitude
+lower than writing directly to Netlink, even though the `svm_fifo` library code has many bells and
+whistles that we don't need. But, perhaps due to these bells and whistles, we may be holding it
+wrong, as invariably after a short while the Netlink writes return _Message too long_ errors.
+
+```
+pim@hvn6-lab:~$ grep 'vector rates' v3fifo-sflow-100-runtime.txt  | grep -v 'in 0'
+  vector rates in 1.0783e7, out 1.0783e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 9.3499e6, out 9.3499e6, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4728e7, out 1.4728e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.3516e7, out 1.3516e7, drop 0.0000e0, punt 0.0000e0
+pim@hvn6-lab:~$ grep -i sflow v3fifo-sflow-100-runtime.txt 
+Name            State        Calls      Vectors  Suspends  Clocks    Vectors/Call  
+sflow           active     1096132    280609792         0  1.63e1          256.00
+sflow           active     1584577    405651712         0  1.46e1          256.00
+pim@hvn6-lab:~$ grep -i sflow v3fifo-sflow-100-err.txt 
+ 280635904               sflow                     sflow packets processed         error  
+   2079194               sflow                      sflow packets sampled          error  
+ 733447310               sflow                    CPU cycles in sent samples       error  
+ 405689856               sflow                     sflow packets processed         error  
+   3004118               sflow                      sflow packets sampled          error  
+1844674407               sflow                    CPU cycles in sent samples       error  
+```
+
+Two things of note here. Firstly, the average clocks spent in the `sFlow` node have gone down from
+86 CPU cycles/packet to 16.3 CPU cycles. But even more importantly, the amount of time spent after
+the sample is taken is hugely reduced, from 1600+ cycles in v2 to a much more favorable 352 cycles
+in this version. Also, any risk of Netlink writes failing has been eliminated, because that's now
+offloaded to a different thread entirely.
+
+**Debrief**: It's not great that we created a new linux `pthread` for the consumer of the samples.
+VPP has an elaborate thread management system, and collaborative multitasking in its threading
+model, which adds introspection like clock counters, names, `show runtime`, `show threads` and so
+on. I can't help but wonder: wouldn't we just be able to move the `spt_process_samples()` thread
+into a VPP process node instead?
+
+### v3bis: SVM FIFO, PSAMPLE process in Main
+
+**TL/DR:** _9.68Mpps L3, 14.10Mpps L2XC, 14.2 cycles/packet, still with corrupted FIFO queue messages_
+
+Neil agrees that there's no good reason to keep this out of main, and conjures up
+[[df2dab8d](https://github.com/vpp/sflow-vpp/df2dab8d)] which rewrites the thread to an
+`sflow_process_samples()` function, using `VLIB_REGISTER_NODE` to add it to VPP in an idiomatic way.
+As a really nice benefit, we can now count how many CPU cycles are spent, in _main_, each time this
+_process_ wakes up and does some work. It's a widely used pattern in VPP.
+
+Because of the FIFO queue message corruption, Netlink message are failing to send at an alarming
+rate, which is causing lots of `clib_warning()` messages to be spewed on console. I replace those
+with a counter of Failed Netlink messages instead, and commit refactor
+[[6ba4715](https://github.com/sflow/vpp-sflow/6ba4715d050f76cfc582055958d50bf4cc8a0ad1)].
+
+```
+pim@hvn6-lab:~$ grep 'vector rates' v3bis-100-runtime.txt  | grep -v 'in 0'
+  vector rates in 1.0976e7, out 1.0976e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 9.6743e6, out 9.6743e6, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4866e7, out 1.4866e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4052e7, out 1.4052e7, drop 0.0000e0, punt 0.0000e0
+pim@hvn6-lab:~$ grep sflow v3bis-100-runtime.txt 
+Name                   State       Calls    Vectors  Suspends  Clocks    Vectors/Call  
+sflow-process-samples  any wait        0          0     28052  4.66e4            0.00
+sflow                  active    1134102  290330112         0  1.42e1          256.00
+sflow                  active    1647240  421693440         0  1.32e1          256.00
+pim@hvn6-lab:~$ grep sflow v3bis-100-err.txt 
+     77945               sflow                        sflow PSAMPLE sent           error  
+       863               sflow                    sflow PSAMPLE send failed        error  
+ 290376960               sflow                     sflow packets processed         error  
+   2151184               sflow                      sflow packets sampled          error  
+ 421761024               sflow                     sflow packets processed         error  
+   3119625               sflow                      sflow packets sampled          error  
+```
+
+With this iteration, I make a few observations. Firstly, the `sflow-process-samples` node shows up
+and informs me that, when handling the samples from the worker FIFO queues, the _process_ is using
+4660 CPU cycles. Secondly, the replacement of `clib_warnign()` with the `sflow PSAMPLE send failed`
+counter reduced time from 16.3 to 14.2 cycles on average in the dataplane. Nice.
+
+**Debrief**: A sad conclusion: of the 5.2M samples taken, only 77k make it through to Netlink. All
+these send failures and corrupt packets are really messing things up. So while the provided FIFO
+implementation in `svm/fifo_segment.h` is idiomatic, it is also much more complex than we thought,
+and we're fearing that it may not be safe to read from another thread.
+
+### v4: Custom lockless FIFO, PSAMPLE process in Main
+
+**TL/DR:** _9.56Mpps L3, 13.69Mpps L2XC, 15.6 cycles/packet, corruption fixed!_
+
+After reading around a bit in DPDK's
+[[kni_fifo](https://doc.dpdk.org/api-18.11/rte__kni__fifo_8h_source.html)], Neil produces a gem of a
+commit in
+[[42bbb64](https://github.com/sflow/vpp-sflow/commit/42bbb643b1f11e8498428d3f7d20cde4de8ee201)],
+where he introduces a tiny multiple-writer, single-consumer FIFO with two simple functions:
+`sflow_fifo_enqueue()` to be called in the workers, and `sflow_fifo_dequeue()` to be called in the
+main thread's `sflow-process-samples` process. He then makes this thread-safe by doing what I
+consider black magic, in commit
+[[dd8af17](https://github.com/sflow/vpp-sflow/commit/dd8af1722d579adc9d08656cd7ec8cf8b9ac11d6)],
+which makes use of `clib_atomic_load_acq_n()` and `clib_atomic_store_rel_n()` macros from VPP's
+`vppinfra/atomics.h`.
+
+What I really like about this change is that it introduces a FIFO implementation in about twenty
+lines of code, which means the sampling code path in the dataplane becomes really easy to follow,
+and will be even faster than it was before. I take it out for a loadtest:
+
+```
+pim@hvn6-lab:~$ grep 'vector rates' v4-100-runtime.txt  | grep -v 'in 0'
+  vector rates in 1.0958e7, out 1.0958e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 9.5633e6, out 9.5633e6, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4849e7, out 1.4849e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.3697e7, out 1.3697e7, drop 0.0000e0, punt 0.0000e0
+pim@hvn6-lab:~$ grep sflow v4-100-runtime.txt 
+Name                   State       Calls    Vectors  Suspends  Clocks    Vectors/Call  
+sflow-process-samples  any wait        0          0     17767  1.52e6            0.00
+sflow                  active    1121156  287015936         0  1.56e1          256.00
+sflow                  active    1605772  411077632         0  1.53e1          256.00
+pim@hvn6-lab:~$ grep sflow v4-100-err.txt 
+   3553600               sflow                        sflow PSAMPLE sent           error  
+ 287101184               sflow                     sflow packets processed         error  
+   2127024               sflow                      sflow packets sampled          error  
+    350224               sflow                      sflow packets dropped          error  
+ 411199744               sflow                     sflow packets processed         error  
+   3043693               sflow                      sflow packets sampled          error  
+   1266893               sflow                      sflow packets dropped          error  
+```
+
+
+This is starting to be a very nice implementation! With this iteration of the plugin, all the
+corruption is gone, there is a slight regression (because we're now actually _sending_ the
+messages). With the v3bis variant, only a tiny fraction of the samples made it through to netlink.
+With this v4 variant, I can see 2127024 + 3043693 packets sampled, but due to a carefully chosen
+FIFO depth of 4, the workers will drop samples so as not to overload the main process that is trying
+to write them out. At this unnatural rate of 1:100, I can see that of the 2127024 samples taken,
+350224 are prematurely dropped (because the FIFO queue is full). This is a perfect defense in depth!
+
+Doing the math, both workers can enqueue 1776800 samples in 30 seconds, which is 59k/s per
+interface. I can also see that the second interface, which is doing L2XC and hits a much larger
+packets/sec throughput, is dropping more samples because it receives an equal amount of time from main
+reading samples from its queue. In other words: in an overload scenario, one interface cannot crowd
+out another. Slick.
+
+Finally, completing my math, each worker has enqueued 1776800 samples to their FIFOs, and I see that
+main has dequeued exactly 2x1776800 = 3553600 samples, all successfully written to Netlink, so
+the `sflow PSAMPLE send failed` counter remains zero.
+
+{{< image float="right" width="20em" src="/assets/sflow/hsflowd-demo.png" >}}
+
+**Debrief**: In the mean time, Neil has been working on the `host-sflow` daemon changes to pick up
+these netlink messages. There's also a bit of work to do with retrieving the packet and byte
+counters of the VPP interfaces, so he is creating a module in `host-sflow` that can consume some
+messages from VPP. He will call this `mod_vpp`, and he mails a screenshot of his work in progress.
+I'll discuss the end-to-end changes with `hsflowd` in a followup article, and focus my efforts here
+on documenting the VPP parts only. But, as a teaser,  here's a screenshot of a validated
+`sflow-tool` output of a VPP instance using our `sFlow` plugin and his pending `host-sflow` changes
+to integrate the rest of the business logic outside of the VPP dataplane, where it's arguably
+expensive to make mistakes.
+
+Neil admits to an itch that he has been meaning to scratch all this time. In VPP's
+`plugins/sflow/node.c`, we insert the node between `device-input` and `ethernet-input`. Here, really
+most of the time the plugin is just shoveling the ethernet packets through to `ethernet-input`. To
+make use of some CPU instruction cache affinity, the loop that does this shovelling can do it one
+packet at a time, two packets at a time, or even four packets at a time. Although the code is super
+repetitive and somewhat ugly, it does actually speed up processing in terms of CPU cycles spent per
+packet, if you shovel four of them at a time.
+
+### v5: Quad Bucket Brigade in worker
+
+**TL/DR:** _9.68Mpps L3, 14.0Mpps L2XC, 11 CPU cycles/packet, 1.28e5 CPU cycles in main_
+
+Neil calls this the _Quad Bucket Brigade_, and one last finishing touch is to move from his default
+2-packet to a 4-packet shoveling. In commit
+[[285d8a0](https://github.com/sflow/vpp-sflow/commit/285d8a097b74bb38eeb14a922a1e8c1115da2ef2)], he
+extends a common pattern in VPP dataplane nodes, each time the node iterates, it'll pre-fetch now up
+to eight packets (`p0-p7`) if the vector is long enough, and handle them four at a time (`b0-b3`).
+He also adds a few compiler hints with branch prediction: almost no packets will have a trace
+enabled, so he can use `PREDICT_FALSE()` macros to allow the compiler to further optimize the code.
+
+I find reading the dataplane code, that it is incredibly ugly. But it's the price to pay for ultra
+fast throughput. But how do we see the effect? My low-tech proposal is to enable sampling at a very
+high rate, say 1:10'000'000, so that the code path that grabs and enqueues the sample into the FIFO
+is almost never called. Then, what's left for the `sFlow` dataplane node, really is to shovel the
+packets from `device-input` into `ethernet-input`.
+
+To measure the relative improvement, I do one test with, and one without commit
+[[285d8a09](https://github.com/sflow/vpp-sflow/commit/285d8a09)].
+
+```
+pim@hvn6-lab:~$ grep 'vector rates' v5-10M-runtime.txt  | grep -v 'in 0'
+  vector rates in 1.0981e7, out 1.0981e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 9.8806e6, out 9.8806e6, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4849e7, out 1.4849e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4328e7, out 1.4328e7, drop 0.0000e0, punt 0.0000e0
+pim@hvn6-lab:~$ grep sflow v5-10M-runtime.txt 
+Name                   State       Calls    Vectors  Suspends  Clocks    Vectors/Call  
+sflow-process-samples  any wait        0          0     28467  9.36e3            0.00
+sflow                  active    1158325  296531200         0  1.09e1          256.00
+sflow                  active    1679742  430013952         0  1.11e1          256.00
+
+pim@hvn6-lab:~$ grep 'vector rates' v5-noquadbrigade-10M-runtime.txt  | grep -v in\ 0
+  vector rates in 1.0959e7, out 1.0959e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 9.7046e6, out 9.7046e6, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4849e7, out 1.4849e7, drop 0.0000e0, punt 0.0000e0
+  vector rates in 1.4008e7, out 1.4008e7, drop 0.0000e0, punt 0.0000e0
+pim@hvn6-lab:~$ grep sflow v5-noquadbrigade-10M-runtime.txt 
+Name                   State       Calls    Vectors  Suspends  Clocks    Vectors/Call  
+sflow-process-samples  any wait        0          0     28462  9.57e3            0.00
+sflow                  active    1137571  291218176         0  1.26e1          256.00
+sflow                  active    1641991  420349696         0  1.20e1          256.00
+```
+
+Would you look at that, this optimization actually works as advertised! There is a meaningful
+_progression_ from v5-noquadbrigade (9.70Mpps L3, 14.00Mpps L2XC) to v5 (9.88Mpps L3, 14.32Mpps
+L2XC).  So at the expense of adding 63 lines of code, there is a 2.8% increase in throughput.
+**Quad-Bucket-Brigade, yaay!**
+
+I'll leave you with a beautiful screenshot of the current code at HEAD, as it is sampling 1:100
+packets (!) on four interfaces, while forwarding 8x10G of 256 byte packets at line rate. You'll
+recall at the beginning of this article I did an acceptance loadtest with sFlow disabled, but this
+is the exact same result **with sFlow** enabled:
+
+{{< image src="/assets/sflow/trex-sflow-acceptance.png" alt="T-Rex sFlow Acceptance Loadtest" >}}
+
+This picture says it all: 79.98 Gbps in, 79.98 Gbps out; 36.22Mpps in, 36.22Mpps out. Also: 176k
+samples/sec taken from the dataplane, with correct rate limiting due to a per-worker FIFO depth
+limit, yielding 25k samples/sec sent to Netlink.
+
+## What's Next
+
+Checking in on the three main things we wanted to ensure with the plugin:
+
+1.  ✅ If `sFlow` _is not_ enabled on a given interface, there is no regression on other interfaces.
+1.  ✅ If `sFlow` _is_ enabled, copying packets costs 11 CPU cycles on average
+1.  ✅ If `sFlow` takes a sample, it takes only marginally more CPU time to enqueue.
+    *   No sampling gets 9.88Mpps of IPv4 and 14.3Mpps of L2XC throughput,
+    *   1:1000 sampling reduces to 9.77Mpps of L3 and 14.05Mpps of L2XC throughput,
+    *   and an overly harsh 1:100 reduces to 9.69Mpps and 13.97Mpps only.
+
+The hard part is finished, but we're not entirely done yet. What's left is to implement a set of
+packet and byte counters, and send this information along with possible Linux CP data (such as the
+TAP interface ID in the Linux side), and to add the module for VPP in `hsflowd`. I'll write about
+that part in a followup article.
+
+Neil has introduced vpp-dev@ to this plugin, and so far there were no objections. But he has pointed
+folks to a github out of tree repo, and I may add a Gerrit instead so it becomes part of the
+ecosystem. Our work so far is captured in Gerrit [[41680](https://gerrit.fd.io/r/c/vpp/+/41680)],
+which ends up being just over 2600 lines all-up. I do think we need to refactor a bit, add some
+VPP-specific tidbits like `FEATURE.yaml` and `*.rst` documentation, but this should be in reasonable
+shape.
+
+### Acknowledgements
+
+I'd like to thank Neil McKee from inMon for his dedication to getting things right, including the
+finer details such as logging, error handling, API specifications, and documentation. He has been a
+true pleasure to work with and learn from.

content/articles/2024-10-21-freeix-2.md

@@ -0,0 +1,778 @@
+---
+date: "2024-10-21T10:52:11Z"
+title: "FreeIX Remote - Part 2"
+---
+
+{{< image width="18em" float="right" src="/assets/freeix/freeix-artist-rendering.png" alt="FreeIX, Artists Rendering" >}}
+
+# Introduction
+
+A few months ago, I wrote about [[an idea]({{< ref 2024-04-27-freeix-1.md >}})] to help boost the
+value of small Internet Exchange Points (_IXPs_). When such an exchange  doesn't have many members,
+then the operational costs of connecting to it (cross connects, router ports, finding peers, etc)
+are not very favorable. 
+
+Clearly, the benefit of using an Internet Exchange is to reduce the portion of an ISP’s (and CDN’s)
+traffic that must be delivered via their upstream transit providers, thereby reducing the average
+per-bit delivery cost and as well reducing the end to end latency as seen by their users or
+customers. Furthermore, the increased number of paths available through the IXP improves routing
+efficiency and fault-tolerance, and at the same time it avoids traffic going the scenic route to a
+large hub like Frankfurt, London, Amsterdam, Paris or Rome, if it could very well remain local.
+
+## Refresher: FreeIX Remote
+
+{{< image width="20em" float="right" src="/assets/freeix/Free IX Remote.svg" alt="FreeIX Remote" >}}
+
+Let's take for example the [[Free IX in Greece](https://free-ix.gr/)] that was announced at GRNOG16
+in Athens on April 19th, 2024.  This exchange initially targets Athens and Thessaloniki, with 2x100G
+between the two cities.  Members can connect to either site for the cost of only a cross connect.
+The 1G/10G/25G ports will be _Gratis_, so please make sure to apply if you're in this region! I
+myself have connected one very special router to Free IX Greece, which will be offering an outreach
+infrastructure by connecting to _other_ Internet Exchange Points in Amsterdam, and allowing all FreeIX
+Greece members to benefit from that in the following way:
+
+1.   FreeIX Remote uses AS50869 to peer with any network operator (or routeserver) available at public
+Internet Exchange Points or using private interconnects. For these peers, it looks like a completely
+normal service provider in this regard. It will connect to internet exchange points, and learn a bunch of
+routes and announce other routes.
+
+1.   FreeIX Remote _members_ can join the program, after which they are granted certain propagation
+permissions by FreeIX Remote at the point where they have a BGP session with AS50869. The prefixes
+learned on these _member_ sessions are marked as such, and will be allowed to propagate.  Members
+will receive some or all learned prefixes from AS50869.
+
+1.   FreeIX _members_ can set fine grained BGP communities to determine which of their prefixes are
+propagated to and from which locations, by router, country or Internet Exchange Point.
+
+Members at smaller internet exchange points greatly benefit from this type of outreach, by receiving large
+portions of the public internet directly at their preferred peering location. The _Free IX Remote_
+routers will carry member traffic to and from these remote Internet Exchange Points.  My [[previous
+article]({{< ref 2024-04-27-freeix-1.md >}})] went into a good amount of detail on the principles of
+operation, but back then I made a promise to come back to the actual _implementation_ of such a
+complex routing topology. As a starting point, I work with the structure I shared in [[IPng's
+Routing Policy]({{< ref 2021-11-14-routing-policy.md >}})]. If you haven't read that yet, I think
+it may make sense to take a look as many of the structural elements and concepts will be similar.
+
+## Implementation
+
+The routing policy calls for three classes of (large) BGP communities: informational, permission and
+inhibit. It also defines a few classic BGP communties, but I'll skip over those as they are not
+very interesting. Firstly, I will use the _informational_ communities to tag which prefixes were
+learned by which _router_, in which _country_ and at which internet exchange point, which I will call a
+_group_. 
+
+Then, I will use the same structure to grant members _permissions_, that is to say, when AS50869
+learns their prefixes, they will get tagged with specific action communities that enable propagation
+to other places. I will call this 'Member-to-IXP'. Sometimes, I'd like to be able to _inhibit_
+propagation of 'Member-to-IXP', so there will be a third set of communities that perform this
+function. Finally, matching on the informational communities in a clever way will enable a symmetric
+'IXP-to-Member' propagation.
+
+To help structure this implementation, it helps if I think about it in
+the following way:
+
+Let's say, AS50869 is connected to IXP1, IXP2, IXP3 and IXP4. AS50869 has a _member_ called M1 at
+IXP1, and that member is 'permitted' to reach IXP2 and IXP3, but it is 'inhibited' from reaching
+IXP4. My _FreeIX Remote_ implementation now has to satisfy three main requirements:
+
+1.  **Ingress**: learn prefixes (from peers and members alike) at internet exchange points or
+private network interconnects, and 'tag' them with the correct informational communities.
+1.  **Egress: Member-to-IXP**: Announce M1's prefixes to IXP2 and IXP3, but not to IXP4.
+1.  **Egress: IXP-to-Member**: Announce IXP2's and IXP3's prefixes to M1, but not IXP4's.
+
+### Defining Countries and Routers
+
+I'll start by giving each country which has at least one router a unique _country_id_ in a YAML
+file, leaving the value 0 to mean 'all' countries:
+
+```
+$ cat config/common/countries.yaml
+country:
+  all: 0
+  CH: 1
+  NL: 2
+  GR: 3
+  IT: 4
+```
+
+Each router has its own configuration file, and at the top, I'll define some metadata which
+includes things like the country in which it operates, and its own unique _router_id_, like so:
+
+```
+$ cat config/chrma0.net.free-ix.net.yaml
+device:
+  id: 1
+  hostname: chrma0.free-ix.net
+  shortname: chrma0
+  country: CH
+  loopbacks:
+    ipv4: 194.126.235.16
+    ipv6: "2a0b:dd80:3101::"
+  location: "Hofwiesenstrasse, Ruemlang, Zurich, Switzerland"
+...
+```
+
+### Defining communities
+
+Next, I define the BGP communities in `class` and `subclass` types, in the following YAML structure:
+
+```
+ebgp:
+  community:
+    legacy:
+      noannounce: 0
+      blackhole: 666
+      inhibit: 3000
+      prepend1: 3100
+      prepend2: 3200
+      prepend3: 3300
+    large:
+      class:
+        informational: 1000
+        permission: 2000
+        inhibit: 3000
+        prepend1: 3100
+        prepend2: 3200
+        prepend3: 3300
+      subclass:
+        all: 0
+        router: 10
+        country: 20
+        group: 30
+        asn: 40
+```
+
+### Defining Members
+
+In order to keep this system manageable, I have to rely on automation. I intend to leverage the
+BGP community _subclasses_ in a simple ACL system consisting of the following YAML, taking my buddy
+Antonios' network as an example:
+
+```
+$ cat config/common/members.yaml
+member:
+  210312:
+    description: DaKnObNET
+    prefix_filter: AS-SET-DNET
+    permission: [ router:chrma0 ]
+    inhibit: [ group:chix ]
+  ...
+```
+
+The syntax of the `permission` and `inhibit` fields are identical. They are lists of key:value pairs
+where they key  must be one of the _subclasses_ (eg. 'router', 'country', 'group', 'asn'), and the
+value appropriate for that type. In this example, AS50869 is being asked to grant permissions for
+Antonios' prefixes to any peer connected to `router:chrma0`, but inhibit propagation to/from the
+exchange point called `group:chix`. I could extend this list, for example by adding a permission to
+`country:NL` or an inhibit to `router:grskg0` and so on.
+
+I decide that sensible defaults are to give permissions to all, and keep inhibit empty. In other
+words: be very liberal in propagation, to maximize the value that FreeIX Remote can provide its
+members.
+
+### Ingress: Learning Prefixes
+
+With what I've defined so far, I can start to set informational BGP communtiies:
+*   The prefixes learned on subclass **router** for `chrma0` will have value of device.id=1:
+`(50869,1010,1)`
+*   The prefixes learned on subclass **country** for `chrma0` will learn from device.country=CH and
+be able to look up in `countries['CH']` that this means value 1: `(50869,1020,1)`
+*   When learning prefixes from a given internet exchange, Kees already knows its PeeringDB
+_ixp_id_, which is a unique value for each exchange point. Thus, subclass **group** for `chrma0` at
+[[CommunityIX](https://www.peeringdb.com/ix/2013)] is ixp_id=2013: `(50869,1030,2013)`
+
+#### Ingress: Learning from members
+
+I need to make sure that members send only the prefixes that I expect from them. To do this, I'll
+make use of a common tool called [[bgpq4](https://github.com/bgp/bgpq4)] which cobbles together the
+prefixes belonging to an AS-SET by referencing one or more IRR databases.
+
+In Python, I'll prepare the Jinja context by generating the prefix filter lists like so:
+
+```
+if session["type"] == "member":
+  session = {**session, **data["member"][asn]}
+
+pf = ebgp_merge_value(data["ebgp"], group, session, "prefix_filter", None)
+if pf: 
+    ctx["prefix_filter"] = {}
+    pfn = pf
+    pfn = pfn.replace("-", "_")
+    pfn = pfn.replace(":", "_")
+
+    for af in [4, 6]:
+        filter_name = "%s_%s_IPV%d" % (groupname.upper(), pfn, af)
+        filter_contents = fetch_bgpq(filter_name, pf, af, allow_morespecifics=True) 
+        if "[" in filter_contents:
+            ctx["prefix_filter"][filter_name] = { "str": filter_contents, "af": af }
+            ctx["prefix_filter_ipv%d" % af] = True
+        else:
+            log.warning(f"Filter {filter_name} is empty!")
+            ctx["prefix_filter_ipv%d" % af] = False
+```
+
+First, if a given BGP session is of type _member_, I'll merge the `member[asn]` dictionary
+into the `ebgp.group.session[asn]`. I've left out error handling for brevity, but in case the member
+YAML file doesn't have an entry for the given ASN, it'll just revert back to being of type _peer_.
+
+I'll use a helper function `ebgp_merge_value()` to walk the YAML hiearchy from the member-data
+enriched _session_ to the _group_ and finally to the _ebgp_ scope, looking for the existence of a
+key called _prefix_filter_ and defaulting to None in case none was found.  With the value of
+_prefix_filter_ in hand (in this case `AS-SET-DNET`), I shell out to `bgpq4` for IPv4 and IPv6
+respectively. Sometimes, there are no IPv6 prefixes (why must you be like this?!) and sometimes
+there are no IPv4 prefixes (welcome to the Internet, kid!)
+
+All of this context, including the session and group information, are then fed as context to a
+Jinja renderer, where I can use them in an _import_ filter like so:
+
+```
+{% for plname, pl in (prefix_filter | default({})).items() %}
+{{pl.str}}
+{% endfor %}
+
+filter ebgp_{{group_name}}_{{their_asn}}_import {
+{% if not prefix_filter_ipv4 | default(True) %}
+  # WARNING: No IPv4 prefix filter found
+  if (net.type = NET_IP4) then reject;
+{% endif %}
+{% if not prefix_filter_ipv6 | default(True) %}
+  # WARNING: No IPv6 prefix filter found
+  if (net.type = NET_IP6) then reject;
+{% endif %}
+{% for plname, pl in (prefix_filter | default({})).items() %}
+{% if pl.af == 4 %}
+  if (net.type = NET_IP4 && ! (net ~ {{plname}})) then reject;
+{% elif pl.af == 6 %}
+  if (net.type = NET_IP6 && ! (net ~ {{plname}})) then reject;
+{% endif %}
+{% endfor %}
+{% if session_type is defined %}
+  if ! ebgp_import_{{session_type}}({{their_asn}}) then reject;
+{% endif %}
+
+  # Add FreeIX Remote: Informational
+  bgp_large_community.add(({{my_asn}},{{community.large.class.informational+community.large.subclass.router}},{{device.id}})); ## informational.router = {{ device.hostname }}
+  bgp_large_community.add(({{my_asn}},{{community.large.class.informational+community.large.subclass.country}},{{country[device.country]}})); ## informational.country = {{ device.country }}
+{% if group.peeringdb_ix.id %}
+  bgp_large_community.add(({{my_asn}},{{community.large.class.informational+community.large.subclass.group}},{{group.peeringdb_ix.id}})); ## informational.group = {{ group_name }}
+{% endif %}
+
+  ## NOTE(pim): More comes here, see Member-to-IXP below
+
+  accept;
+}
+```
+
+Let me explain what's going on here, as Jinja templating language that my generator uses is a bit
+... chatty. The first block will print the dictionary of zero or more `prefix_filter` entries. If
+the `prefix_filter` context variable doesn't exist, assume it's the empty dictionary and thus,
+print no prefix lists.
+
+Then, I create a Bird2 filter and these must each have a globally unique name. I satisfy this
+requirement by giving it a name with the tuple of {group, their_asn}. The first thing this filter
+does, is inspect `prefix_filter_ipv4` and `prefix_filter_ipv6`, and if they are explicitly set to
+False (for example, if a member doesn't have any IRR prefixes associated with their AS-SET), then
+I'll reject any prefixes from them. Then, I'll match the prefixes with the `prefix_filter`, if
+provided, and reject any prefixes that aren't in the list I'm expecting on this session. Assuming
+we're still good to go, I'll hand this prefix off to a function called `ebgp_import_peer()` for
+peers and `ebgp_import_member()` for members, both of which ensure BGP communities are scrubbed.
+
+```
+function ebgp_import_peer(int remote_as) -> bool
+{
+  # Scrub BGP Communities (RFC 7454 Section 11)
+  bgp_community.delete([(50869, *)]);
+  bgp_large_community.delete([(50869, *, *)]);
+
+  # Scrub BLACKHOLE community
+  bgp_community.delete((65535, 666));
+
+  return ebgp_import(remote_as);
+}
+
+function ebgp_import_member(int remote_as) -> bool
+{
+  # We scrub only our own (informational, permissions) BGP Communities for members
+  bgp_large_community.delete([(50869,1000..2999,*)]);
+
+  return ebgp_import(remote_as);
+}
+```
+
+After scrubbing the communities (peers are not allowed to set _any_ communities, and members are not
+allowed to set their own informational or permissions communities, but they are allowed to inhibit
+themselves or prepend, if they wish), one last check is performed by calling the underlying
+`ebgp_import()`:
+
+```
+function ebgp_import(int remote_as) -> bool
+{
+  if aspath_bogon() then return false;
+  if (net.type = NET_IP4 && ipv4_bogon()) then return false;
+  if (net.type = NET_IP6 && ipv6_bogon()) then return false;
+
+  if (net.type = NET_IP4 && ipv4_rpki_invalid()) then return false;
+  if (net.type = NET_IP6 && ipv6_rpki_invalid()) then return false;
+
+  # Graceful Shutdown (https://www.rfc-editor.org/rfc/rfc8326.html)
+  if (65535, 0) ~ bgp_community then bgp_local_pref = 0;
+
+  return true;
+}
+```
+
+Here, belt-and-suspenders checks are performed, notably bogon AS Paths, IPv4/IPv6 prefixes and RPKI
+invalids are filtered out. If the prefix has well-known community for [[BGP Graceful
+Shutdown](https://www.rfc-editor.org/rfc/rfc8326.html)], honor it and set the local preference to
+zero (making sure to prefer any other available path).
+
+OK, after all these checks are done, I am finally ready to accept the prefix from this peer or
+member. It's time to add the informational communities based on the _router_id_, the router's
+_country_id_ and (if this is a session at a public internet exchange point documented in PeeringDB),
+the group's _ixp_id_.
+
+#### Ingress Example: member
+
+Here's what the rendered template looks like for Antonios' member session at CHIX:
+
+```
+# bgpq4 -Ab4 -R 32 -l 'define CHIX_AS_SET_DNET_IPV4' AS-SET-DNET
+define CHIX_AS_SET_DNET_IPV4 = [
+ 44.31.27.0/24{24,32}, 44.154.130.0/24{24,32}, 44.154.132.0/24{24,32},
+ 147.189.216.0/21{21,32}, 193.5.16.0/22{22,32}, 212.46.55.0/24{24,32}
+];
+
+# bgpq4 -Ab6 -R 128 -l 'define CHIX_AS_SET_DNET_IPV6' AS-SET-DNET
+define CHIX_AS_SET_DNET_IPV6 = [
+ 2001:678:f5c::/48{48,128}, 2a05:dfc1:9174::/48{48,128}, 2a06:9f81:2500::/40{40,128},
+ 2a06:9f81:2600::/40{40,128}, 2a0a:6044:7100::/40{40,128}, 2a0c:2f04:100::/40{40,128},
+ 2a0d:3dc0::/29{29,128}, 2a12:bc0::/29{29,128}
+];
+
+filter ebgp_chix_210312_import {
+  if (net.type = NET_IP4 && ! (net ~ CHIX_AS_SET_DNET_IPV4)) then reject;
+  if (net.type = NET_IP6 && ! (net ~ CHIX_AS_SET_DNET_IPV6)) then reject;
+  if ! ebgp_import_member(210312) then reject;
+
+  # Add FreeIX Remote: Informational
+  bgp_large_community.add((50869,1010,1)); ## informational.router = chrma0.free-ix.net
+  bgp_large_community.add((50869,1020,1)); ## informational.country = CH
+  bgp_large_community.add((50869,1030,2365)); ## informational.group = chix
+  
+  ## NOTE(pim): More comes here, see Member-to-IXP below
+
+  accept;
+}
+```
+
+#### Ingress Example: peer
+
+For completeness, here's a regular peer Cloudflare at CHIX, and I hope you agree that the Jinja
+template renders down to something waaaay more readable now:
+
+```
+filter ebgp_chix_13335_import {
+  if ! ebgp_import_peer(13335) then reject;
+
+  # Add FreeIX Remote: Informational
+  bgp_large_community.add((50869,1010,1)); ## informational.router = chrma0.free-ix.net
+  bgp_large_community.add((50869,1020,1)); ## informational.country = CH
+  bgp_large_community.add((50869,1030,2365)); ## informational.group = chix
+
+  accept;
+}
+```
+
+Most sessions will actually look like this one: just learning prefixes, scrubbing inbound
+communities that are nobody's business to be setting but mine, tossing weird prefixes like bogons
+and then setting typically the three informational communities. I now know exactly which prefixes
+are picked up at group CHIX, which ones in country Switzerland, and which ones on router `chrma0`.
+
+### Egress: Propagating Prefixes
+
+And with that, I've completed the 'learning' part. Let me move to the 'propagating' part. A design
+goal of FreeIX Remote is to have _symmetric_ propagation. In my example above, member M1 should have
+its prefixes announced at IXP2 and IXP3, and all prefixes learned at IXP2 and IXP3 should be
+announced to member M1.
+
+First, let me create a helper function in the generator. It's job is to take the symbolic
+`member.*.permissions` and `member.*.inhibit` lists and resolve them into a structure of numeric
+values suitable for BGP community list adding and matching. It's a bit of a beast, but I've
+simplified it a bit. Notably, I've removed all the error and exception handling for brevity:
+
+```
+def parse_member_communities(data, asn, type):
+  myasn = data["ebgp"]["asn"]
+  cls = data["ebgp"]["community"]["large"]["class"]
+  sub = data["ebgp"]["community"]["large"]["subclass"]
+
+  bgp_cl = []
+  member = data["member"][asn]
+
+  for perm in perms:
+    if perm == "all":
+      el = { "class": int(cls[type]), "subclass": int(sub["all"]),
+             "value": 0, "description": f"{type}.all" }
+      return [el]
+    k, v = perm.split(":")
+    if k == "country":
+      country_id = data["country"][v]
+      el = { "class": int(cls[type]), "subclass": int(sub["country"]),
+             "value": int(country_id), "description": f"{type}.{k} = {v}" }
+      bgp_cl.append(el)
+    elif k == "asn":
+      el = { "class": int(cls[type]), "subclass": int(sub["asn"]),
+             "value": int(v), "description": f"{type}.{k} = {v}" }
+      bgp_cl.append(el)
+    elif k == "router":
+      device_id = data["_devices"][v]["id"]
+      el = { "class": int(cls[type]), "subclass": int(sub["router"]),
+             "value": int(device_id), "description": f"{type}.{k} = {v}" }
+      bgp_cl.append(el)
+    elif k == "group":
+      group = data["ebgp"]["groups"][v]
+      if isinstance(group["peeringdb_ix"], dict):
+        ix_id = group["peeringdb_ix"]["id"]
+      else:
+        ix_id = group["peeringdb_ix"]
+      el = { "class": int(cls[type]), "subclass": int(sub["group"]),
+             "value": int(ix_id), "description": f"{type}.{k} = {v}" }
+      bgp_cl.append(el)
+    else:
+      log.warning (f"No implementation for {type} subclass '{k}' for member AS{asn}, skipping")
+
+    return bgp_cl
+
+```
+
+The essence of this function is to take a human readable list of symbols, like 'router:chrma0' and
+look up what subclass is called 'router' and what router_id is 'chrma0'. It does this for keywords
+'router', 'country', 'group' and 'asn' and for a special keyword called 'all' as well.
+
+Running this a function on Antonios' member data above would reveal the following:
+```
+Member 210312 has permissions:
+ [{'class': 2000, 'subclass': 10, 'value': 1, 'description': 'permission.router = chrma0'}]
+Member 210312 has inhibits:
+ [{'class': 3000, 'subclass': 30, 'value': 2365, 'description': 'inhibit.group = chix'}]
+```
+
+The neat thing about this is, that this data will come in handy for _both_ types of propagation, and
+the `parse_member_communities()` helper function returns pretty readable data, which will help in
+debugging and further understanding the ultimately generated configuration.
+
+#### Egress: Member-to-IXP
+
+OK, when I learned Antonios' prefixes, I have instructed the system to propagate them to all
+sessions on router `chrma0`, except sessions on group `chix`. This means that in the direction of
+_from AS50869 to others_, I can do the following:
+
+**1. Tag permissions and inhibits on ingress**
+
+I add a tiny bit of logic using this data structure I just created above. In the import filter,
+remember I added `NOTE(pim): More comes here`? After setting the informational communities, I also
+add these:
+
+```
+{% if session_type == "member" %}
+{% if permissions %}
+
+  # Add FreeIX Remote: Permission
+{% for el in permissions %}
+  bgp_large_community.add(({{my_asn}},{{el.class+el.subclass}},{{el.value}})); ## {{ el.description
+}}
+{% endfor %}
+{% endif %}
+{% if inhibits %}
+
+  # Add FreeIX Remote: Inhibit
+{% for el in inhibits %}
+  bgp_large_community.add(({{my_asn}},{{el.class+el.subclass}},{{el.value}})); ## {{ el.description
+}}
+{% endfor %}
+{% endif %}
+{% endif %}
+```
+
+Seeing as this block only gets rendered if the session type is _member_, let me show you how
+Antonios' import filter looks like in its full glory:
+
+```
+filter ebgp_chix_210312_import {
+  if (net.type = NET_IP4 && ! (net ~ CHIX_AS_SET_DNET_IPV4)) then reject;
+  if (net.type = NET_IP6 && ! (net ~ CHIX_AS_SET_DNET_IPV6)) then reject;
+  if ! ebgp_import_member(210312) then reject;
+
+  # Add FreeIX Remote: Informational
+  bgp_large_community.add((50869,1010,1)); ## informational.router = chrma0.free-ix.net
+  bgp_large_community.add((50869,1020,1)); ## informational.country = CH
+  bgp_large_community.add((50869,1030,2365)); ## informational.group = chix
+
+  # Add FreeIX Remote: Permission
+  bgp_large_community.add((50869,2010,1)); ## permission.router = chrma0
+
+  # Add FreeIX Remote: Inhibit
+  bgp_large_community.add((50869,3030,2365)); ## inhibit.group = chix
+
+  accept;
+}
+```
+
+Remember, the `ebgp_import_member()` helper will strip any informational (the 1000s) and permissions
+(the 2000s), but it would allow Antonios to set inhibits and prepends (the 3000s) so these BGP
+communities will still be allowed in. In other words, Antonios can't give himself propagation rights
+(sorry, buddy!) but if he would like to make AS50869 stop sending his prefixes to, say, CommunityIX,
+he could simply add the BGP community `(50869,3030,2013)` on his announcements, and that will get
+honored. If he'd like AS50869 to prepend itself twice before announcing to peer AS8298, he could set
+`(50869,3200,8298)` and that will also get picked up.
+
+**2. Match permissions and inhibits on egress**
+
+Now that all of Antonios' prefixes are tagged with permissions and inhibits, I can reveal how I
+implemented the export filters for AS50869:
+
+```
+function member_prefix(int group) -> bool
+{ 
+  bool permitted = false;
+
+  if (({{ebgp.asn}}, {{ebgp.community.large.class.permission+ebgp.community.large.subclass.all}}, 0) ~ bgp_large_community ||
+      ({{ebgp.asn}}, {{ebgp.community.large.class.permission+ebgp.community.large.subclass.router}}, {{ device.id }}) ~ bgp_large_community ||
+      ({{ebgp.asn}}, {{ebgp.community.large.class.permission+ebgp.community.large.subclass.country}}, {{ country[device.country] }}) ~ bgp_large_community ||
+      ({{ebgp.asn}}, {{ebgp.community.large.class.permission+ebgp.community.large.subclass.group}}, group) ~ bgp_large_community) then {
+    permitted = true;
+  }
+  if (({{ebgp.asn}}, {{ebgp.community.large.class.inhibit+ebgp.community.large.subclass.all}}, 0) ~ bgp_large_community ||
+      ({{ebgp.asn}}, {{ebgp.community.large.class.inhibit+ebgp.community.large.subclass.router}}, {{ device.id }}) ~ bgp_large_community ||
+      ({{ebgp.asn}}, {{ebgp.community.large.class.inhibit+ebgp.community.large.subclass.country}}, {{ country[device.country] }}) ~ bgp_large_community ||
+      ({{ebgp.asn}}, {{ebgp.community.large.class.inhibit+ebgp.community.large.subclass.group}}, group) ~ bgp_large_community) then {
+    permitted = false;
+  }
+  return (permitted);
+}
+
+function valid_prefix(int group) -> bool
+{    
+  return (source_prefix() || member_prefix(group));
+}    
+
+function ebgp_export_peer(int remote_as; int group) -> bool
+{
+  if (source != RTS_BGP && source != RTS_STATIC) then return false;
+  if !valid_prefix(group) then return false;
+
+  bgp_community.delete([(50869, *)]);
+  bgp_large_community.delete([(50869, *, *)]);
+
+  return ebgp_export(remote_as);
+}
+```
+
+From the bottom, the function `ebgp_export_peer()` is invoked on each peering session, and it gets
+the argument of the remote AS (for example 13335 for CloudFlare), and the group (for example 2365
+for CHIX). The function ensures that it's either a _static_ route or a _BGP_ route. Then it makes
+sure it's a `valid_prefix()` for the group.
+
+The `valid_prefix()` function first checks if it's one of our own (as in: AS50869's own) prefixes,
+which it does by calling `source_prefix()`, which i've ommitted here as it would be a distraction.
+All it does is check if the prefix is in a static prefix list generated with `bgpq4` for AS50869
+itself. The more interesting observation is that to be eligible, the prefix needs to be either
+`source_prefix()` **or** `member_prefix(group)`.
+
+The propagation decision for 'Member-to-IXP' actually happens in that `member_prefix()` function. It
+starts off by assuming the prefix is not permitted. Then it scans all relevant _permissions_
+communities which may be present in the RIB for this prefix:
+- is the `all` permissions community `(50869,2000,0)` set?
+- what about the `router` permission `(50869,2010,R)` for my _router_id_?
+- perhaps the `country` permission `(50869,2020,C)` for my _country_id_?
+- or maybe the `group` permission `(50869,2030,G)` for the _ixp_id_ that this session lives on?
+
+If any of these conditions are true, then this prefix _might_ pe permitted, so I set the variable to
+True. Next, I check and see if any of the _inhibit_ communities are set, either by me (in
+`members.yaml`) or by the member on the live BGP session. If any one of them  matches, then I flip
+the variable to False again. Once the verdict is known, I can return True or False here, which
+makes its way all the way up the call stack and ultimately announces the member prefix on the BGP
+session, or not. Slick!
+
+#### Egress: IXP-to-Member
+
+At this point, members' prefixes get announced at the correct internet exchange points, but I need to
+satisfy one more requirement: the prefixes picked up at those IXPs, should _also_ be announced to
+members. For this, the helper dictionary with permissions and inhibits can be used in a clever way.
+What if I held them against the informational communities? For example, I have _permitted_
+Antonios to be annouced at any IXP connected to router `chrma0`, then all prefixes I learned at
+`chrma0` are fair game, right? But, I configured an _inhibit_ for Antonios' prefixes at CHIX. No
+problem, I have an informational community for all prefixes I learned from the CHIX group!
+
+I come to the realization that IXP-to-Member simply adds to the Member-to-IXP logic. Everything that
+I would announce to a peer, I will also announce to a member. Off I go, adding one last helper
+function to the BGP session Jinja template:
+
+```
+{% if session_type == "member" %}
+function ebgp_export_{{group_name}}_{{their_asn}}(int remote_as; int group) -> bool
+{
+  bool permitted = false;
+
+  if (source != RTS_BGP && source != RTS_STATIC) then return false;
+  if valid_prefix(group) then return ebgp_export(remote_as);
+
+{% for el in permissions | default([]) %}
+  if (bgp_large_community ~ [({{ my_asn }},{{ 1000+el.subclass}},{% if el.value == 0%}*{% else %}{{el.value}}{% endif %})]) then permitted=true; ## {{el.description}}
+{% endfor %}
+{% for el in inhibits | default([]) %}
+  if (bgp_large_community ~ [({{ my_asn }},{{ 1000+el.subclass}},{% if el.value == 0%}*{% else %}{{el.value}}{% endif %})]) then permitted=false; ## {{el.description}}
+{% endfor %}
+
+  if (permitted) then return ebgp_export(remote_as);
+  return false;
+}
+{% endif %}
+```
+
+Note that in essence, this new function still calls `valid_prefix()`, which in turn calls
+`source_prefix()` **or** `member_prefix(group)`, so it announces the same prefixes that are also
+announced to sessions of type 'peer'. But then, I'll also inspect the _informational_ communities,
+where the value of `0` is replaced with a wildcard, because 'permit or inhibit all' would mean
+'match any of these BGP communities'. This template renders as follows for Antonios at CHIX:
+
+```
+function ebgp_export_chix_210312(int remote_as; int group) -> bool
+{
+  bool export = false;
+
+  if (source != RTS_BGP && source != RTS_STATIC) then return false;
+  if valid_prefix(group) then return ebgp_export(remote_as);
+
+  if (bgp_large_community ~ [(50869,1010,1)]) then export=true; ## permission.router = chrma0
+  if (bgp_large_community ~ [(50869,1030,2365)]) then export=false; ## inhibit.group = chix
+
+  if (export) then return ebgp_export(remote_as);
+  return false;
+}
+```
+
+## Results
+
+With this, the propagation logic is complete. Announcements are _symmetric_, that is to say the function
+`ebgp_export_chix_210312()` sees to it that Antonios gets the prefixes learned at router `chrma0`
+but not those learned at group `CHIX`. Similarly, the `ebgp_export_peer()` ensures that Antonios'
+prefixes are propagated to any session at router `chrma0` except those sessions at group `CHIX`.
+
+{{< image width="200px" float="right" src="/assets/vpp/fdio-color.svg" alt="VPP" >}}
+
+I have installed VPP with [[OSPFv3]({{< ref 2024-06-22-vpp-ospf-2.md >}})] unnumbered interfaces,
+so each router has exactly one IPv4 and IPv6 loopback address. The router in R&uuml;mlang has been
+operational for a while, the one in Amsterdam (nlams0.free-ix.net) and Thessaloniki
+(grskg0.free-ix.net) have been deployed and are connecting to IXPs now, and the one in Milan
+(itmil0.free-ix.net) has been installed but is pending physical deployment at Caldara.
+
+I deployed a test setup with a few permissions and inhibits on the R&uuml;mlang router, with many thanks
+to Jurrian, Sam and Antonios for allowing me to guinnaepig-ize their member sessions. With the
+following test configuration:
+
+```
+member:
+  35202:
+    description: OnTheGo (Sam Aschwanden)
+    prefix_filter: AS-OTG
+    permission: [ router:chrma0 ]
+    inhibit: [ group:comix ]
+  210312:
+    description: DaKnObNET
+    prefix_filter: AS-SET-DNET
+    permission: [ router:chrma0 ]
+    inhibit: [ group:chix ]
+  212635:
+    description: Jurrian van Iersel
+    prefix_filter: AS212635:AS-212635
+    permission: [ router:chrma0 ]
+    inhibit: [ group:chix, group:fogixp ]
+```
+
+I can see the following prefix learn/announce counts towards _members_:
+
+```
+pim@chrma0:~$ for i in $(birdc show protocol | grep member | cut -f1 -d' '); do echo -n $i\ ; birdc
+show protocol all $i | grep Routes; done
+chix_member_35202_ipv4_1        2 imported, 0 filtered, 159984 exported, 0 preferred
+chix_member_35202_ipv6_1        2 imported, 0 filtered, 61730 exported, 0 preferred
+chix_member_210312_ipv4_1       3 imported, 0 filtered, 3518 exported, 3 preferred
+chix_member_210312_ipv6_1       2 imported, 0 filtered, 1251 exported, 2 preferred
+comix_member_35202_ipv4_1       2 imported, 0 filtered, 159981 exported, 2 preferred
+comix_member_35202_ipv4_2       2 imported, 0 filtered, 159981 exported, 1 preferred
+comix_member_35202_ipv6_1       2 imported, 0 filtered, 61727 exported, 2 preferred
+comix_member_35202_ipv6_2       2 imported, 0 filtered, 61727 exported, 1 preferred
+fogixp_member_212635_ipv4_1     1 imported, 0 filtered, 442 exported, 1 preferred
+fogixp_member_212635_ipv6_1     14 imported, 0 filtered, 181 exported, 14 preferred
+freeix_ch_member_210312_ipv4_1  3 imported, 0 filtered, 3521 exported, 0 preferred
+freeix_ch_member_210312_ipv6_1  2 imported, 0 filtered, 1253 exported, 0 preferred
+```
+
+Let me make a few observations:
+*   Hurricane Electric AS6939 is present at CHIX, and they tend to announce a very large number of
+prefixes. So every member who is permitted (and not inhibited) at CHIX will see all of those: Sam's
+AS35202 is inhibited on CommunityIX but not on CHIX, and he's permitted on both. That explains why
+he is seeing the routes on both sessions.
+*   I've inhibited Jurrian's AS212635 to/from both CHIX and FogIXP, which means he will be seeing
+CommunityIX (~245 IPv4, 85 IPv6 prefixes), and FreeIX CH (~173 IPv4 and ~60 IPv6). We also send him
+the member prefixes, which is about 35 or so additional prefixes. This explains why Jurrian is
+receiving from us ~440 IPv4 and ~180 IPv6.
+*   Antonios' AS210312, the exemplar in this article, is receiving all-but-CHIX. FogIXP yields 3077
+or so IPv4 and 1056 IPv6 prefixes, while I've already added up FreeIX, CommunityIX, and our members
+(this is what we're sending Jurrian!), at 330 resp 180, so Antonios should be getting about 3500 IPv4
+prefixes and 1250 IPv6 prefixes.
+
+In the other direction, I would expect to be announcing to _peers_ only prefixes belonging to either
+AS50869 itself, or those of our members:
+
+```
+pim@chrma0:~$ for i in $(birdc show protocol | grep peer.*_1 | cut -f1 -d' '); do echo -n $i\ ; birdc
+show protocol all $i | grep Routes || echo; done
+chix_peer_212100_ipv4_1      57618 imported, 0 filtered, 24 exported, 778 preferred
+chix_peer_212100_ipv6_1      21979 imported, 1 filtered, 37 exported, 7186 preferred
+chix_peer_13335_ipv4_1       4767 imported, 9 filtered, 24 exported, 4765 preferred
+chix_peer_13335_ipv6_1       371 imported, 1 filtered, 37 exported, 369 preferred
+chix_peer_6939_ipv4_1        151787 imported, 27 filtered, 24 exported, 133943 preferred
+chix_peer_6939_ipv6_1        61191 imported, 6 filtered, 37 exported, 16223 preferred
+comix_peer_44596_ipv4_1      594 imported, 0 filtered, 25 exported, 10 preferred
+comix_peer_44596_ipv6_1      1147 imported, 0 filtered, 50 exported, 0 preferred
+comix_peer_8298_ipv4_1       23 imported, 0 filtered, 25 exported, 0 preferred
+comix_peer_8298_ipv6_1       34 imported, 0 filtered, 50 exported, 0 preferred
+fogixp_peer_47498_ipv4_1     3286 imported, 1 filtered, 27 exported, 3077 preferred
+fogixp_peer_47498_ipv6_1     1838 imported, 0 filtered, 39 exported, 1056 preferred
+freeix_ch_peer_51530_ipv4_1  355 imported, 0 filtered, 28 exported, 0 preferred
+freeix_ch_peer_51530_ipv6_1  143 imported, 0 filtered, 53 exported, 0 preferred
+```
+
+Some observations:
+
+*   Nobody is inhibited at FreeIX Switzerland. It stands to reason therefore, that it has the most
+exported prefixes: 28 for IPv4 and 53 for IPv6.
+*   Two members are inhibited at CHIX, which makes it have the lowest amount of exported prefixes:
+24 for IPv4 and 27 for IPv6.
+*   All members at each exchange (group) will have the same amount of prefixes. I can confirm that
+at CHIX, all thre peers have the same amount of announced prefixes. Similarly, at CommunityIX, all
+peers have the same amount.
+*   If Antonios, Sam or Jurrian would add an outgoing announcement to AS50869 with an additional inhibit
+BGP community (eg `(50869,3020,1)` to inhibit country Switzerland), they could tweak these numbers.
+
+## What's next
+
+This all adds up. I'd like to test the waters with my friendly neighborhood canaries a little bit,
+to make sure that announcements are expected, and traffic flows where appropriate. In the mean time,
+I'll chase the deployment of LSIX, FrysIX, SpeedIX and possibly a few others in Amsterdam. And of
+course FreeIX Greece in Thessaloniki. I'll try to get the Milano VPP router deployed (it's already
+installed and configured, but currently powered off) and connected to PCIX, MIX and a few others.
+
+## How can you help?
+
+If you're willing to participate with a VPP router and connect it to either multiple local internet
+exchanges (like I've demonstrated in Zurich), or better yet, to one or more of the other existing
+routers, I would welcome your contribution. [[Contact]({{< ref contact.md >}})] me for details.
+
+A bit further down the pike, a connection from Amsterdam to Zurich, from Zurich to Milan and from
+Milan to Thessaloniki is on the horizon. If you are willing and able to donate some bandwidth (point
+to point VPWS, VLL, L2VPN) and your transport network is capable of at least 2026 bytes of _inner_
+payload, please also [[reach out]({{< ref contact.md >}})] as I'm sure many small network operators
+would be thrilled.

content/articles/2025-02-08-sflow-3.md

@@ -0,0 +1,857 @@
+---
+date: "2025-02-08T07:51:23Z"
+title: 'VPP with sFlow - Part 3'
+---
+
+# Introduction
+
+{{< image float="right" src="/assets/sflow/sflow.gif" alt="sFlow Logo" width="12em" >}}
+
+In the second half of last year, I picked up a project together with Neil McKee of
+[[inMon](https://inmon.com/)], the care takers of [[sFlow](https://sflow.org)]: an industry standard
+technology for monitoring high speed networks. `sFlow` gives complete visibility into the
+use of networks enabling performance optimization, accounting/billing for usage, and defense against
+security threats.
+
+The open source software dataplane [[VPP](https://fd.io)] is a perfect match for sampling, as it
+forwards packets at very high rates using underlying libraries like [[DPDK](https://dpdk.org/)] and
+[[RDMA](https://en.wikipedia.org/wiki/Remote_direct_memory_access)]. A clever design choice in the
+so called _Host sFlow Daemon_ [[host-sflow](https://github.com/sflow/host-sflow)], which allows for
+a small portion of code to _grab_ the samples, for example in a merchant silicon ASIC or FPGA, but
+also in the VPP software dataplane. The agent then _transmits_ these samples using a Linux kernel
+feature called [[PSAMPLE](https://github.com/torvalds/linux/blob/master/net/psample/psample.c)].
+This greatly reduces the complexity of code to be implemented in the forwarding path, while at the
+same time bringing consistency to the `sFlow` delivery pipeline by (re)using the `hsflowd` business
+logic for the more complex state keeping, packet marshalling and transmission from the _Agent_ to a
+central _Collector_.
+
+In this third article, I wanted to spend some time discussing how samples make their way out of the
+VPP dataplane, and into higher level tools.
+
+## Recap: sFlow
+
+{{< image float="left" src="/assets/sflow/sflow-overview.png" alt="sFlow Overview" width="14em" >}}
+
+sFlow describes a method for Monitoring Traffic in Switched/Routed Networks, originally described in
+[[RFC3176](https://datatracker.ietf.org/doc/html/rfc3176)]. The current specification is version 5
+and is homed on the sFlow.org website [[ref](https://sflow.org/sflow_version_5.txt)]. Typically, a
+Switching ASIC in the dataplane (seen at the bottom of the diagram to the left) is asked to copy
+1-in-N packets to local sFlow Agent.
+
+**Sampling**: The agent will copy the first N bytes (typically 128) of the packet into a sample. As
+the ASIC knows which interface the packet was received on, the `inIfIndex` will be added. After a
+routing decision is made, the nexthop and its L2 address and interface become known. The ASIC might
+annotate the sample with this `outIfIndex` and `DstMAC` metadata as well.
+
+**Drop Monitoring**: There's one rather clever insight that sFlow gives: what if the packet _was
+not_ routed or switched, but rather discarded?  For this, sFlow is able to describe the reason for
+the drop. For example, the ASIC receive queue could have been overfull, or it did not find a
+destination to forward the packet to (no FIB entry), perhaps it was instructed by an ACL to drop the
+packet or maybe even tried to transmit the packet but the physical datalink layer had to abandon the
+transmission for whatever reason (link down, TX queue full, link saturation, and so on). It's hard
+to overstate how important it is to have this so-called _drop monitoring_, as operators often spend
+hours and hours figuring out _why_ packets are lost their network or datacenter switching fabric.
+
+**Metadata**: The agent may have other metadata as well, such as which prefix was the source and
+destination of the packet, what additional RIB information is available (AS path, BGP communities,
+and so on).  This may be added to the sample record as well.
+
+**Counters**: Since sFlow is sampling 1:N packets, the system can estimate total traffic in a
+reasonably accurate way. Peter and Sonia wrote a succint
+[[paper](https://sflow.org/packetSamplingBasics/)] about the math, so I won't get into that here.
+Mostly because I am but a software engineer, not a statistician... :) However, I will say this: if a
+fraction of the traffic is sampled but the _Agent_ knows how many bytes and packets were forwarded
+in total, it can provide an overview with a quantifiable accuracy. This is why the _Agent_ will
+periodically get the interface counters from the ASIC.
+
+**Collector**: One or more samples can be concatenated into UDP messages that go from the _sFlow
+Agent_ to a central _sFlow Collector_. The heavy lifting in analysis is done upstream from the
+switch or router, which is great for performance.  Many thousands or even tens of thousands of
+agents can forward their samples and interface counters to a single central collector, which in turn
+can be used to draw up a near real time picture of the state of traffic through even the largest of
+ISP networks or datacenter switch fabrics.
+
+In sFlow parlance [[VPP](https://fd.io/)] and its companion
+[[hsflowd](https://github.com/sflow/host-sflow)] together form an _Agent_ (it sends the UDP packets
+over the network), and for example the commandline tool `sflowtool` could be a _Collector_ (it
+receives the UDP packets).
+
+## Recap: sFlow in VPP
+
+First, I have some pretty good news to report - our work on this plugin was
+[[merged](https://gerrit.fd.io/r/c/vpp/+/41680)] and will be included in the VPP 25.02 release in a
+few weeks! Last weekend, I gave a lightning talk at
+[[FOSDEM](https://fosdem.org/2025/schedule/event/fosdem-2025-4196-vpp-monitoring-100gbps-with-sflow/)]
+in Brussels, Belgium, and caught up with a lot of community members and network- and software
+engineers. I had a great time.
+
+In trying to keep the amount of code as small as possible, and therefore the probability of bugs that
+might impact VPP's dataplane stability low, the architecture of the end to end solution consists of
+three distinct parts, each with their own risk and performance profile:
+
+{{< image float="left" src="/assets/sflow/sflow-vpp-overview.png" alt="sFlow VPP Overview" width="18em" >}}
+
+**1. sFlow worker node**: Its job is to do what the ASIC does in the hardware case. As VPP moves
+packets from `device-input` to the `ethernet-input` nodes in its forwarding graph, the sFlow plugin
+will inspect 1-in-N, taking a sample for further processing. Here, we don't try to be clever, simply
+copy the `inIfIndex` and the first bytes of the ethernet frame, and append them to a
+[[FIFO](https://en.wikipedia.org/wiki/FIFO_(computing_and_electronics))] queue. If too many samples
+arrive, samples are dropped at the tail, and a counter incremented.  This way, I can tell when the
+dataplane is congested. Bounded FIFOs also provide fairness: it allows for each VPP worker thread to
+get their fair share of samples into the Agent's hands.
+
+**2. sFlow main process**: There's a function running on the _main thread_, which shifts further
+processing time _away_ from the dataplane. This _sflow-process_ does two things. Firstly, it
+consumes samples from the per-worker FIFO queues (both forwarded packets in green, and dropped ones
+in red). Secondly, it keeps track of time and every few seconds (20 by default, but this is
+configurable), it'll grab all interface counters from those interfaces for which I have sFlow
+turned on. VPP produces _Netlink_ messages and sends them to the kernel.
+
+**3. Host sFlow daemon**: The third component is external to VPP: `hsflowd` subscribes to the _Netlink_
+messages. It goes without saying that `hsflowd` is a battle-hardened implementation running on
+hundreds of different silicon and software defined networking stacks. The PSAMPLE stuff is easy,
+this module already exists. But Neil implemented a _mod_vpp_ which can grab interface names and their
+`ifIndex`, and counter statistics. VPP emits this data as _Netlink_ `USERSOCK` messages alongside
+the PSAMPLEs.
+
+
+By the way, I've written about _Netlink_ before when discussing the [[Linux Control Plane]({{< ref
+2021-08-25-vpp-4 >}})] plugin.  It's a mechanism for programs running in userspace to share
+information with the kernel. In the Linux kernel, packets can be sampled as well, and sent from
+kernel to userspace using a _PSAMPLE_ Netlink channel. However, the pattern is that of a message
+producer/subscriber relationship and nothing precludes one userspace process (`vpp`) to be the
+producer while another userspace process (`hsflowd`) acts as the consumer!
+
+Assuming the sFlow plugin in VPP produces samples and counters properly, `hsflowd` will do the rest,
+giving correctness and upstream interoperability pretty much for free. That's slick!
+
+### VPP: sFlow Configuration
+
+The solution that I offer is based on two moving parts. First, the VPP plugin configuration, which
+turns on sampling at a given rate on physical devices, also known as _hardware-interfaces_. Second,
+the open source component [[host-sflow](https://github.com/sflow/host-sflow/releases)] can be
+configured as of release v2.11-5 [[ref](https://github.com/sflow/host-sflow/tree/v2.1.11-5)].
+
+I will show how to configure VPP in three ways:
+
+***1. VPP Configuration via CLI***
+
+```
+pim@vpp0-0:~$ vppctl
+vpp0-0# sflow sampling-rate 100
+vpp0-0# sflow polling-interval 10
+vpp0-0# sflow header-bytes 128
+vpp0-0# sflow enable GigabitEthernet10/0/0
+vpp0-0# sflow enable GigabitEthernet10/0/0 disable
+vpp0-0# sflow enable GigabitEthernet10/0/2
+vpp0-0# sflow enable GigabitEthernet10/0/3
+```
+
+The first three commands set the global defaults - in my case I'm going to be sampling at 1:100
+which is an unusually high rate. A production setup may take 1-in-_linkspeed-in-megabits_ so for a
+1Gbps device 1:1'000 is appropriate. For 100GE, something between 1:10'000 and 1:100'000 is more
+appropriate, depending on link load. The second command sets the interface stats polling interval.
+The default is to gather these statistics every 20 seconds, but I set it to 10s here.
+
+Next, I tell the plugin how many bytes of the sampled ethernet frame should be taken. Common
+values are 64 and 128 but it doesn't have to be a power of two. I want enough data to see the
+headers, like MPLS label(s), Dot1Q tag(s), IP header and TCP/UDP/ICMP header, but the contents of
+the payload are rarely interesting for
+statistics purposes.
+
+Finally, I can turn on the sFlow plugin on an interface with the `sflow enable-disable` CLI. In VPP,
+an idiomatic way to turn on and off things is to have an enabler/disabler. It feels a bit clunky
+maybe to write `sflow enable $iface disable` but it makes more logical sends if you parse that as
+"enable-disable" with the default being the "enable" operation, and the alternate being the
+"disable" operation.
+
+***2. VPP Configuration via API***
+
+I implemented a few API methods for the most common operations. Here's a snippet that obtains the
+same config as what I typed on the CLI above, but using these Python API calls:
+
+```python
+from vpp_papi import VPPApiClient, VPPApiJSONFiles
+import sys
+
+vpp_api_dir = VPPApiJSONFiles.find_api_dir([])
+vpp_api_files = VPPApiJSONFiles.find_api_files(api_dir=vpp_api_dir)
+vpp = VPPApiClient(apifiles=vpp_api_files, server_address="/run/vpp/api.sock")
+vpp.connect("sflow-api-client")
+print(vpp.api.show_version().version)
+# Output: 25.06-rc0~14-g9b1c16039
+
+vpp.api.sflow_sampling_rate_set(sampling_N=100)
+print(vpp.api.sflow_sampling_rate_get())
+# Output: sflow_sampling_rate_get_reply(_0=655, context=3, sampling_N=100)
+
+vpp.api.sflow_polling_interval_set(polling_S=10)
+print(vpp.api.sflow_polling_interval_get())
+# Output: sflow_polling_interval_get_reply(_0=661, context=5, polling_S=10)
+
+vpp.api.sflow_header_bytes_set(header_B=128)
+print(vpp.api.sflow_header_bytes_get())
+# Output: sflow_header_bytes_get_reply(_0=665, context=7, header_B=128)
+
+vpp.api.sflow_enable_disable(hw_if_index=1, enable_disable=True)
+vpp.api.sflow_enable_disable(hw_if_index=2, enable_disable=True)
+print(vpp.api.sflow_interface_dump())
+# Output: [ sflow_interface_details(_0=667, context=8, hw_if_index=1),
+#           sflow_interface_details(_0=667, context=8, hw_if_index=2) ]
+
+print(vpp.api.sflow_interface_dump(hw_if_index=2))
+# Output: [ sflow_interface_details(_0=667, context=9, hw_if_index=2) ]
+
+print(vpp.api.sflow_interface_dump(hw_if_index=1234)) ## Invalid hw_if_index
+# Output: []
+
+vpp.api.sflow_enable_disable(hw_if_index=1, enable_disable=False)
+print(vpp.api.sflow_interface_dump())
+# Output: [ sflow_interface_details(_0=667, context=10, hw_if_index=2) ]
+```
+
+This short program toys around a bit with the sFlow API. I first set the sampling to 1:100 and get
+the current value. Then I set the polling interval to 10s and retrieve the current value again.
+Finally, I set the header bytes to 128, and retrieve the value again. 
+
+Enabling and disabling sFlow on interfaces shows the idiom I mentioned before - the API being an
+`*_enable_disable()` call of sorts, and typically taking a boolean argument if the operator wants to
+enable (the default), or disable sFlow on the interface. Getting the list of enabled interfaces can
+be done with the `sflow_interface_dump()` call, which returns a list of `sflow_interface_details`
+messages.
+
+I demonstrated VPP's Python API and how it works in a fair amount of detail in a [[previous
+article]({{< ref 2024-01-27-vpp-papi >}})], in case this type of stuff interests you.
+
+***3. VPPCfg YAML Configuration***
+
+Writing on the CLI and calling the API is good and all, but many users of VPP have noticed that it
+does not have any form of configuration persistence and that's deliberate. VPP's goal is to be a
+programmable dataplane, and explicitly has left the programming and configuration as an exercise for
+integrators. I have written a Python project that takes a YAML file as input and uses it to
+configure (and reconfigure, on the fly) the dataplane automatically, called
+[[VPPcfg](https://git.ipng.ch/ipng/vppcfg.git)]. Previously, I wrote some implementation thoughts
+on its [[datamodel]({{< ref 2022-03-27-vppcfg-1 >}})] and its [[operations]({{< ref 2022-04-02-vppcfg-2
+>}})] so I won't repeat that here. Instead, I will just show the configuration:
+
+```
+pim@vpp0-0:~$ cat << EOF > vppcfg.yaml
+interfaces:
+  GigabitEthernet10/0/0:
+    sflow: true
+  GigabitEthernet10/0/1:
+    sflow: true
+  GigabitEthernet10/0/2:
+    sflow: true
+  GigabitEthernet10/0/3:
+    sflow: true
+
+sflow:
+  sampling-rate: 100
+  polling-interval: 10
+  header-bytes: 128
+EOF
+pim@vpp0-0:~$ vppcfg plan -c vppcfg.yaml -o /etc/vpp/config/vppcfg.vpp
+[INFO    ] root.main: Loading configfile vppcfg.yaml
+[INFO    ] vppcfg.config.valid_config: Configuration validated successfully
+[INFO    ] root.main: Configuration is valid
+[INFO    ] vppcfg.reconciler.write: Wrote 13 lines to /etc/vpp/config/vppcfg.vpp
+[INFO    ] root.main: Planning succeeded
+pim@vpp0-0:~$ vppctl exec /etc/vpp/config/vppcfg.vpp
+```
+
+The nifty thing about `vppcfg` is that if I were to change, say, the sampling-rate (setting it to
+1000) and disable sFlow from an interface, say Gi10/0/0, I can re-run the `vppcfg plan` and `vppcfg
+apply` stages and the VPP dataplane will be reprogrammed to reflect the newly declared configuration.
+
+### hsflowd: Configuration
+
+When sFlow is enabled, VPP will start to emit _Netlink_ messages of type PSAMPLE with packet samples
+and of type USERSOCK with the custom messages containing interface names and counters. These latter
+custom messages have to be decoded, which is done by the _mod_vpp_ module in `hsflowd`, starting
+from release v2.11-5 [[ref](https://github.com/sflow/host-sflow/tree/v2.1.11-5)].
+
+Here's a minimalist configuration:
+
+```
+pim@vpp0-0:~$ cat /etc/hsflowd.conf
+sflow {
+  collector { ip=127.0.0.1 udpport=16343 }
+  collector { ip=192.0.2.1 namespace=dataplane }
+  psample { group=1 }
+  vpp { osIndex=off }
+}
+```
+
+{{< image width="5em" float="left" src="/assets/shared/warning.png" alt="Warning" >}}
+
+There are two important details that can be confusing at first: \
+**1.** kernel network namespaces \
+**2.** interface index namespaces
+
+#### hsflowd: Network namespace
+
+Network namespaces virtualize Linux's network stack. Upon creation, a network namespace contains only
+a loopback interface, and subsequently interfaces can be moved between namespaces.  Each network
+namespace will have its own set of IP addresses, its own routing table, socket listing, connection
+tracking table, firewall, and other network-related resources.  When started by systemd, `hsflowd`
+and VPP will normally both run in the _default_ network namespace.
+
+Given this, I can conclude that when the sFlow plugin opens a Netlink channel, it will
+naturally do this in the network namespace that its VPP process is running in (the _default_
+namespace, normally). It is therefore important that the recipient of these Netlink messages,
+notably `hsflowd` runs in ths ***same*** namespace as VPP. It's totally fine to run them together in
+a different namespace (eg. a container in Kubernetes or Docker), as long as they can see each other.
+
+It might pose a problem if the network connectivity lives in a different namespace than the default
+one. One common example (that I heavily rely on at IPng!) is to create Linux Control Plane interface
+pairs, _LIPs_, in a dataplane namespace. The main reason for doing this is to allow something like
+FRR or Bird to completely govern the routing table in the kernel and keep it in-sync with the FIB in
+VPP. In such a _dataplane_ network namespace, typically every interface is owned by VPP.
+
+Luckily, `hsflowd` can attach to one (default) namespace to get the PSAMPLEs, but create a socket in
+a _different_ (dataplane) namespace to send packets to a collector. This explains the second
+_collector_ entry in the config-file above. Here, `hsflowd` will send UDP packets to 192.0.2.1:6343
+from within the (VPP) dataplane namespace, and to 127.0.0.1:16343 in the default namespace.
+
+#### hsflowd: osIndex
+
+I hope the previous section made some sense, because this one will be a tad more esoteric. When
+creating a network namespace, each interface will get its own uint32 interface index that identifies
+it, and such an ID is typically called an `ifIndex`. It's important to note that the same number can
+(and will!) occur multiple times, once for each namespace. Let me give you an example:
+
+```
+pim@summer:~$ ip link
+1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN ...
+    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
+2: eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc mq master ipng-sl state UP ...
+    link/ether 00:22:19:6a:46:2e brd ff:ff:ff:ff:ff:ff
+    altname enp1s0f0
+3: eno2: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 900 qdisc mq master ipng-sl state DOWN ...
+    link/ether 00:22:19:6a:46:30 brd ff:ff:ff:ff:ff:ff
+    altname enp1s0f1
+
+pim@summer:~$ ip netns exec dataplane ip link
+1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN ...
+    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
+2: loop0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9216 qdisc mq state UP ...
+    link/ether de:ad:00:00:00:00 brd ff:ff:ff:ff:ff:ff
+3: xe1-0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9216 qdisc mq state UP ...
+    link/ether 00:1b:21:bd:c7:18 brd ff:ff:ff:ff:ff:ff
+```
+
+I want to draw your attention to the number at the beginning of the line. In the _default_
+namespace, `ifIndex=3` corresponds to `ifName=eno2` (which has no link, it's marked `DOWN`). But in
+the _dataplane_ namespace, that index corresponds to a completely different interface called
+`ifName=xe1-0` (which is link `UP`).
+
+Now, let me show you the interfaces in VPP:
+
+```
+pim@summer:~$ vppctl show int | grep Gigabit | egrep 'Name|loop0|tap0|Gigabit'
+              Name               Idx    State  MTU (L3/IP4/IP6/MPLS)
+GigabitEthernet4/0/0              1      up          9000/0/0/0
+GigabitEthernet4/0/1              2     down         9000/0/0/0
+GigabitEthernet4/0/2              3     down         9000/0/0/0
+GigabitEthernet4/0/3              4     down         9000/0/0/0
+TenGigabitEthernet5/0/0           5      up          9216/0/0/0
+TenGigabitEthernet5/0/1           6      up          9216/0/0/0
+loop0                             7      up          9216/0/0/0
+tap0                              19     up          9216/0/0/0
+```
+
+Here, I want you to look at the second column `Idx`, which shows what VPP calls the _sw_if_index_
+(the software interface index, as opposed to hardware index). Here, `ifIndex=3` corresponds to
+`ifName=GigabitEthernet4/0/2`, which is neither `eno2` nor `xe1-0`. Oh my, yet _another_ namespace!
+
+It turns out that there are three (relevant) types of namespaces at play here:
+1.  ***Linux network*** namespace; here using `dataplane` and `default` each with their own unique
+    (and overlapping) numbering.
+1.  ***VPP hardware*** interface namespace, also called PHYs (for physical interfaces). When VPP
+    first attaches to or creates network interfaces like the ones from DPDK or RDMA, these will
+    create an _hw_if_index_ in a list.
+1.  ***VPP software*** interface namespace. All interfaces (including hardware ones!) will
+    receive a _sw_if_index_ in VPP. A good example is sub-interfaces: if I create a sub-int on
+    GigabitEthernet4/0/2, it will NOT get a hardware index, but it _will_ get the next available
+    software index (in this example, `sw_if_index=7`).
+
+In Linux CP, I can see a mapping from one to the other, just look at this:
+
+```
+pim@summer:~$ vppctl show lcp
+lcp default netns dataplane
+lcp lcp-auto-subint off
+lcp lcp-sync on
+lcp lcp-sync-unnumbered on
+itf-pair: [0] loop0 tap0 loop0 2 type tap netns dataplane
+itf-pair: [1] TenGigabitEthernet5/0/0 tap1 xe1-0 3 type tap netns dataplane
+itf-pair: [2] TenGigabitEthernet5/0/1 tap2 xe1-1 4 type tap netns dataplane
+itf-pair: [3] TenGigabitEthernet5/0/0.20 tap1.20 xe1-0.20 5 type tap netns dataplane
+```
+
+Those `itf-pair` describe our _LIPs_, and they have the coordinates to three things. 1) The VPP
+software interface (VPP `ifName=loop0` with `sw_if_index=7`), which 2) Linux CP will mirror into the
+Linux kernel using a TAP device (VPP `ifName=tap0` with `sw_if_index=19`). That TAP has one leg in
+VPP (`tap0`), and another in 3) Linux (with `ifName=loop` and `ifIndex=2` in namespace `dataplane`).
+
+> So the tuple that fully describes a _LIP_ is `{7, 19,'dataplane', 2}`
+
+Climbing back out of that rabbit hole, I am now finally ready to explain the feature. When sFlow in
+VPP takes its sample, it will be doing this on a PHY, that is a given interface with a specific
+_hw_if_index_. When it polls the counters, it'll do it for that specific _hw_if_index_. It now has a
+choice: should it share with the world the representation of *its* namespace, or should it try to be
+smarter? If LinuxCP is enabled, this interface will likely have a representation in Linux. So the
+plugin will first resolve the _sw_if_index_ belonging to that PHY, and using that, try to look up a
+_LIP_ with it. If it finds one, it'll know both the namespace in which it lives as well as the
+osIndex in that namespace. If it doesn't find a _LIP_, it will at least have the _sw_if_index_ at
+hand, so it'll annotate the USERSOCK counter messages with this information instead.
+
+Now, `hsflowd` has a choice to make: does it share the Linux representation and hide VPP as an
+implementation detail? Or does it share the VPP dataplane _sw_if_index_? There are use cases
+relevant to both, so the decision was to let the operator decide, by setting `osIndex` either `on`
+(use Linux ifIndex) or `off` (use VPP _sw_if_index_).
+
+### hsflowd: Host Counters
+
+Now that I understand the configuration parts of VPP and `hsflowd`, I decide to configure everything
+but without enabling sFlow on on any interfaces yet in VPP. Once I start the daemon, I can see that
+it sends an UDP packet every 30 seconds to the configured _collector_:
+
+```
+pim@vpp0-0:~$ sudo tcpdump -s 9000 -i lo -n
+tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
+listening on lo, link-type EN10MB (Ethernet), snapshot length 9000 bytes
+15:34:19.695042 IP 127.0.0.1.48753 > 127.0.0.1.6343: sFlowv5,
+   IPv4 agent 198.19.5.16, agent-id 100000, length 716
+```
+
+The `tcpdump` I have on my Debian bookworm machines doesn't know how to decode the contents of these
+sFlow packets. Actually, neither does Wireshark. I've attached a file of these mysterious packets
+[[sflow-host.pcap](/assets/sflow/sflow-host.pcap)] in case you want to take a look.
+Neil however gives me a tip. A full message decoder and otherwise handy Swiss army knife lives in
+[[sflowtool](https://github.com/sflow/sflowtool)].
+
+I can offer this pcap file to `sflowtool`, or let it just listen on the UDP port directly, and
+it'll tell me what it finds:
+
+```
+pim@vpp0-0:~$ sflowtool -p 6343      
+startDatagram =================================
+datagramSourceIP 127.0.0.1
+datagramSize 716       
+unixSecondsUTC 1739112018
+localtime 2025-02-09T15:40:18+0100
+datagramVersion 5            
+agentSubId 100000   
+agent 198.19.5.16        
+packetSequenceNo 57 
+sysUpTime 987398   
+samplesInPacket 1            
+startSample ----------------------
+sampleType_tag 0:4                                                         
+sampleType COUNTERSSAMPLE
+sampleSequenceNo 33
+sourceId 2:1                
+counterBlock_tag 0:2001           
+adaptor_0_ifIndex 2                                                        
+adaptor_0_MACs 1   
+adaptor_0_MAC_0 525400f00100
+counterBlock_tag 0:2010
+udpInDatagrams 123904
+udpNoPorts 23132459
+udpInErrors 0
+udpOutDatagrams 46480629
+udpRcvbufErrors 0
+udpSndbufErrors 0
+udpInCsumErrors 0
+counterBlock_tag 0:2009
+tcpRtoAlgorithm 1
+tcpRtoMin 200
+tcpRtoMax 120000
+tcpMaxConn 4294967295
+tcpActiveOpens 0
+tcpPassiveOpens 30
+tcpAttemptFails 0
+tcpEstabResets 0
+tcpCurrEstab 1
+tcpInSegs 89120
+tcpOutSegs 86961
+tcpRetransSegs 59
+tcpInErrs 0
+tcpOutRsts 4
+tcpInCsumErrors 0
+counterBlock_tag 0:2008
+icmpInMsgs 23129314
+icmpInErrors 32
+icmpInDestUnreachs 0
+icmpInTimeExcds 23129282
+icmpInParamProbs 0
+icmpInSrcQuenchs 0
+icmpInRedirects 0
+icmpInEchos 0
+icmpInEchoReps 32
+icmpInTimestamps 0
+icmpInAddrMasks 0
+icmpInAddrMaskReps 0
+icmpOutMsgs 0
+icmpOutErrors 0
+icmpOutDestUnreachs 23132467
+icmpOutTimeExcds 0
+icmpOutParamProbs 23132467
+icmpOutSrcQuenchs 0
+icmpOutRedirects 0
+icmpOutEchos 0
+icmpOutEchoReps 0
+icmpOutTimestamps 0
+icmpOutTimestampReps 0
+icmpOutAddrMasks 0
+icmpOutAddrMaskReps 0
+counterBlock_tag 0:2007
+ipForwarding 2
+ipDefaultTTL 64
+ipInReceives 46590552
+ipInHdrErrors 0
+ipInAddrErrors 0
+ipForwDatagrams 0
+ipInUnknownProtos 0
+ipInDiscards 0
+ipInDelivers 46402357
+ipOutRequests 69613096
+ipOutDiscards 0
+ipOutNoRoutes 80
+ipReasmTimeout 0
+ipReasmReqds 0
+ipReasmOKs 0
+ipReasmFails 0
+ipFragOKs 0
+ipFragFails 0
+ipFragCreates 0
+counterBlock_tag 0:2005
+disk_total 6253608960
+disk_free 2719039488
+disk_partition_max_used 56.52
+disk_reads 11512
+disk_bytes_read 626214912
+disk_read_time 48469
+disk_writes 1058955
+disk_bytes_written 8924332032
+disk_write_time 7954804
+counterBlock_tag 0:2004
+mem_total 8326963200
+mem_free 5063872512
+mem_shared 0
+mem_buffers 86425600
+mem_cached 827752448
+swap_total 0
+swap_free 0
+page_in 306365
+page_out 4357584
+swap_in 0
+swap_out 0
+counterBlock_tag 0:2003
+cpu_load_one 0.030
+cpu_load_five 0.050
+cpu_load_fifteen 0.040
+cpu_proc_run 1
+cpu_proc_total 138
+cpu_num 2
+cpu_speed 1699
+cpu_uptime 1699306
+cpu_user 64269210
+cpu_nice 1810
+cpu_system 34690140
+cpu_idle 3234293560
+cpu_wio 3568580
+cpuintr 0
+cpu_sintr 5687680
+cpuinterrupts 1596621688
+cpu_contexts 3246142972
+cpu_steal 329520
+cpu_guest 0
+cpu_guest_nice 0
+counterBlock_tag 0:2006
+nio_bytes_in 250283
+nio_pkts_in 2931
+nio_errs_in 0
+nio_drops_in 0
+nio_bytes_out 370244
+nio_pkts_out 1640
+nio_errs_out 0
+nio_drops_out 0
+counterBlock_tag 0:2000
+hostname vpp0-0
+UUID ec933791-d6af-7a93-3b8d-aab1a46d6faa
+machine_type 3
+os_name 2
+os_release 6.1.0-26-amd64
+endSample   ----------------------
+endDatagram   =================================
+```
+
+If you thought: "What an obnoxiously long paste!", then my slightly RSI-induced mouse-hand might
+agree with you. But it is really cool to see that every 30 seconds, the _collector_ will receive
+this form of heartbeat from the _agent_. There's a lot of vitalsigns in this packet, including some
+non-obvious but interesting stats like CPU load, memory, disk use and disk IO, and kernel version
+information. It's super dope!
+
+### hsflowd: Interface Counters
+
+Next, I'll enable sFlow in VPP on all four interfaces (Gi10/0/0-Gi10/0/3), set the sampling rate to
+something very high (1 in 100M), and the interface polling-interval to every 10 seconds. And indeed,
+every ten seconds or so I get a few packets, which I captured in
+[[sflow-interface.pcap](/assets/sflow/sflow-interface.pcap)]. Most of the packets contain only one
+counter record, while some contain more than one (in the PCAP, packet #9 has two). If I update the
+polling-interval to every second, I can see that most of the packets have all four counters.
+
+Those interface counters, as decoded by `sflowtool`, look like this:
+
+```
+pim@vpp0-0:~$ sflowtool -r sflow-interface.pcap | \
+              awk '/startSample/ { on=1 } { if (on) { print $0 } } /endSample/ { on=0 }'
+startSample   ----------------------
+sampleType_tag 0:4
+sampleType COUNTERSSAMPLE
+sampleSequenceNo 745
+sourceId 0:3
+counterBlock_tag 0:1005
+ifName GigabitEthernet10/0/2
+counterBlock_tag 0:1
+ifIndex 3
+networkType 6
+ifSpeed 0
+ifDirection 1
+ifStatus 3
+ifInOctets 858282015
+ifInUcastPkts 780540
+ifInMulticastPkts 0
+ifInBroadcastPkts 0
+ifInDiscards 0
+ifInErrors 0
+ifInUnknownProtos 0
+ifOutOctets 1246716016
+ifOutUcastPkts 975772
+ifOutMulticastPkts 0
+ifOutBroadcastPkts 0
+ifOutDiscards 127
+ifOutErrors 28
+ifPromiscuousMode 0
+endSample   ----------------------
+```
+
+What I find particularly cool about it, is that sFlow provides an automatic mapping between the
+`ifName=GigabitEthernet10/0/2` (tag 0:1005), together with an object (tag 0:1), which contains the
+`ifIndex=3`, and lots of packet and octet counters both in the ingress and egress direction. This is
+super useful for upstream _collectors_, as they can now find the hostname, agent name and address,
+and the correlation between interface names and their indexes. Noice!
+
+#### hsflowd: Packet Samples
+
+Now it's time to ratchet up the packet sampling, so I move it from 1:100M to 1:1000, while keeping
+the interface polling-interval at 10 seconds and I ask VPP to sample 64 bytes of each packet that it
+inspects. On either side of my pet VPP instance, I start an `iperf3` run to generate some traffic. I
+now see a healthy stream of sflow packets coming in on port 6343. They still contain every 30
+seconds or so a host counter, and every 10 seconds a set of interface counters come by, but mostly
+these UDP packets are showing me samples. I've captured a few minutes of these in
+[[sflow-all.pcap](/assets/sflow/sflow-all.pcap)].
+Although Wireshark doesn't know how to interpret the sFlow counter messages, it _does_ know how to
+interpret the sFlow sample messagess, and it reveals one of them like this:
+
+{{< image width="100%" src="/assets/sflow/sflow-wireshark.png" alt="sFlow Wireshark" >}}
+
+Let me take a look at the picture from top to bottom. First, the outer header (from 127.0.0.1:48753
+to 127.0.0.1:6343) is the sFlow agent sending to the collector. The agent identifies itself as
+having IPv4 address 198.19.5.16 with ID 100000 and an uptime of 1h52m. Then, it says it's going to
+send 9 samples, the first of which says it's from ifIndex=2 and at a sampling rate of 1:1000. It
+then shows that sample, saying that the frame length is 1518 bytes, and the first 64 bytes of those
+are sampled.  Finally, the first sampled packet starts at the blue line. It shows the SrcMAC and
+DstMAC, and that it was a TCP packet from 192.168.10.17:51028 to 192.168.10.33:5201 - my running
+`iperf3`, booyah!
+
+### VPP: sFlow Performance
+
+{{< image float="right" src="/assets/sflow/sflow-lab.png" alt="sFlow Lab" width="20em" >}}
+
+One question I get a lot about this plugin is: what is the performance impact when using
+sFlow? I spent a considerable amount of time tinkering with this, and together with Neil bringing
+the plugin to what we both agree is the most efficient use of CPU. We could have gone a bit further,
+but that would require somewhat intrusive changes to VPP's internals and as _North of the Border_
+(and the Simpsons!) would say: what we have isn't just good, it's good enough!
+
+I've built a small testbed based on two Dell R730 machines. On the left, I have a Debian machine
+running Cisco T-Rex using four quad-tengig network cards, the classic Intel i710-DA4. On the right,
+I have my VPP machine called _Hippo_ (because it's always hungry for packets), with the same
+hardware.  I'll build two halves. On the top NIC (Te3/0/0-3 in VPP), I will install IPv4 and MPLS
+forwarding on the purple circuit, and a simple Layer2 cross connect on the cyan circuit. On all four
+interfaces, I will enable sFlow. Then, I will mirror this configuration on the bottom NIC
+(Te130/0/0-3) in the red and green circuits, for which I will leave sFlow turned off.
+
+To help you reproduce my results, and under the assumption that this is your jam, here's the
+configuration for all of the kit:
+
+***0. Cisco T-Rex***
+```
+pim@trex:~ $ cat /srv/trex/8x10.yaml
+- version: 2
+  interfaces: [ '06:00.0', '06:00.1', '83:00.0', '83:00.1', '87:00.0', '87:00.1', '85:00.0', '85:00.1' ]
+  port_info:
+    - src_mac:  00:1b:21:06:00:00
+      dest_mac: 9c:69:b4:61:a1:dc    # Connected to Hippo Te3/0/0, purple
+    - src_mac:  00:1b:21:06:00:01
+      dest_mac: 9c:69:b4:61:a1:dd    # Connected to Hippo Te3/0/1, purple
+    - src_mac:  00:1b:21:83:00:00
+      dest_mac: 00:1b:21:83:00:01    # L2XC via Hippo Te3/0/2, cyan
+    - src_mac:  00:1b:21:83:00:01
+      dest_mac: 00:1b:21:83:00:00    # L2XC via Hippo Te3/0/3, cyan
+
+    - src_mac:  00:1b:21:87:00:00
+      dest_mac: 9c:69:b4:61:75:d0    # Connected to Hippo Te130/0/0, red
+    - src_mac:  00:1b:21:87:00:01
+      dest_mac: 9c:69:b4:61:75:d1    # Connected to Hippo Te130/0/1, red
+    - src_mac:  9c:69:b4:85:00:00
+      dest_mac: 9c:69:b4:85:00:01    # L2XC via Hippo Te130/0/2, green
+    - src_mac:  9c:69:b4:85:00:01
+      dest_mac: 9c:69:b4:85:00:00    # L2XC via Hippo Te130/0/3, green
+pim@trex:~ $ sudo t-rex-64 -i -c 4 --cfg /srv/trex/8x10.yaml
+```
+
+When constructing the T-Rex configuration, I specifically set the destination MAC address for L3
+circuits (the purple and red ones) using Hippo's interface MAC address, which I can find with
+`vppctl show hardware-interfaces`. This way, T-Rex does not have to ARP for the VPP endpoint. On
+L2XC circuits (the cyan and green ones), VPP does not concern itself with the MAC addressing at
+all. It puts its interface in _promiscuous_ mode, and simply writes out any ethernet frame received,
+directly to the egress interface.
+
+***1. IPv4***
+```
+hippo# set int state TenGigabitEthernet3/0/0 up
+hippo# set int state TenGigabitEthernet3/0/1 up
+hippo# set int state TenGigabitEthernet130/0/0 up
+hippo# set int state TenGigabitEthernet130/0/1 up
+hippo# set int ip address TenGigabitEthernet3/0/0 100.64.0.1/31
+hippo# set int ip address TenGigabitEthernet3/0/1 100.64.1.1/31
+hippo# set int ip address TenGigabitEthernet130/0/0 100.64.4.1/31
+hippo# set int ip address TenGigabitEthernet130/0/1 100.64.5.1/31
+hippo# ip route add 16.0.0.0/24 via 100.64.0.0
+hippo# ip route add 48.0.0.0/24 via 100.64.1.0
+hippo# ip route add 16.0.2.0/24 via 100.64.4.0
+hippo# ip route add 48.0.2.0/24 via 100.64.5.0
+hippo# ip neighbor TenGigabitEthernet3/0/0 100.64.0.0 00:1b:21:06:00:00 static
+hippo# ip neighbor TenGigabitEthernet3/0/1 100.64.1.0 00:1b:21:06:00:01 static
+hippo# ip neighbor TenGigabitEthernet130/0/0 100.64.4.0 00:1b:21:87:00:00 static
+hippo# ip neighbor TenGigabitEthernet130/0/1 100.64.5.0 00:1b:21:87:00:01 static
+```
+
+By the way, one note to this last piece, I'm setting static IPv4 neighbors so that Cisco T-Rex 
+as well as VPP do not have to use ARP to resolve each other. You'll see above that the T-Rex
+configuration also uses MAC addresses exclusively. Setting the `ip neighbor` like this allows VPP
+to know where to send return traffic.
+
+***2. MPLS***
+```
+hippo# mpls table add 0
+hippo# set interface mpls TenGigabitEthernet3/0/0 enable
+hippo# set interface mpls TenGigabitEthernet3/0/1 enable
+hippo# set interface mpls TenGigabitEthernet130/0/0 enable
+hippo# set interface mpls TenGigabitEthernet130/0/1 enable
+hippo# mpls local-label add 16 eos via 100.64.1.0 TenGigabitEthernet3/0/1 out-labels 17
+hippo# mpls local-label add 17 eos via 100.64.0.0 TenGigabitEthernet3/0/0 out-labels 16
+hippo# mpls local-label add 20 eos via 100.64.5.0 TenGigabitEthernet130/0/1 out-labels 21
+hippo# mpls local-label add 21 eos via 100.64.4.0 TenGigabitEthernet130/0/0 out-labels 20
+```
+
+Here, the MPLS configuration implements a simple P-router, where incoming MPLS packets with label 16
+will be sent back to T-Rex on Te3/0/1 to the specified IPv4 nexthop (for which I already know the
+MAC address), and with label 16 removed and new label 17 imposed, in other words a SWAP operation.
+
+***3. L2XC***
+```
+hippo# set int state TenGigabitEthernet3/0/2 up
+hippo# set int state TenGigabitEthernet3/0/3 up
+hippo# set int state TenGigabitEthernet130/0/2 up
+hippo# set int state TenGigabitEthernet130/0/3 up
+hippo# set int l2 xconnect TenGigabitEthernet3/0/2 TenGigabitEthernet3/0/3
+hippo# set int l2 xconnect TenGigabitEthernet3/0/3 TenGigabitEthernet3/0/2
+hippo# set int l2 xconnect TenGigabitEthernet130/0/2 TenGigabitEthernet130/0/3
+hippo# set int l2 xconnect TenGigabitEthernet130/0/3 TenGigabitEthernet130/0/2
+```
+
+I've added a layer2 cross connect as well because it's computationally very cheap for VPP to receive
+an L2 (ethernet) datagram, and immediately transmit it on another interface. There's no FIB lookup
+and not even an L2 nexthop lookup involved, VPP is just shoveling ethernet packets in-and-out as
+fast as it can!
+
+Here's how a loadtest looks like when sending 80Gbps at 192b packets on all eight interfaces:
+
+{{< image src="/assets/sflow/sflow-lab-trex.png" alt="sFlow T-Rex" width="100%" >}}
+
+The leftmost ports p0 <-> p1 are sending IPv4+MPLS, while ports p0 <-> p2 are sending ethernet back
+and forth. All four of them have sFlow enabled, at a sampling rate of 1:10'000, the default. These
+four ports are my experiment, to show the CPU use of sFlow. Then, ports p3 <-> p4 and p5 <-> p6
+respectively have sFlow turned off but with the same configuration.  They are my control, showing
+the CPU use without sFlow.
+
+**First conclusion**: This stuff works a treat. There is absolutely no impact of throughput at
+80Gbps with 47.6Mpps either _with_, or _without_ sFlow turned on. That's wonderful news, as it shows
+that the dataplane has more CPU available than is needed for any combination of functionality.
+
+But what _is_ the limit? For this, I'll take a deeper look at the runtime statistics by varying the
+CPU time spent and maximum throughput achievable on a single VPP worker, thus using a single CPU
+thread on this Hippo machine that has 44 cores and 44 hyperthreads. I switch the loadtester to emit
+64 byte ethernet packets, the smallest I'm allowed to send.
+
+| Loadtest | no sFlow | 1:1'000'000 | 1:10'000 | 1:1'000 | 1:100 |
+|-------------|-----------|-----------|-----------|-----------|-----------|
+| L2XC        | 14.88Mpps | 14.32Mpps | 14.31Mpps | 14.27Mpps | 14.15Mpps |
+| IPv4        | 10.89Mpps | 9.88Mpps  | 9.88Mpps  | 9.84Mpps  | 9.73Mpps  |
+| MPLS        | 10.11Mpps | 9.52Mpps  | 9.52Mpps  | 9.51Mpps  | 9.45Mpps  |
+| ***sFlow Packets*** / 10sec | N/A       | 337.42M total | 337.39M total | 336.48M total | 333.64M total |
+| .. Sampled    | &nbsp;    | 328 | 33.8k | 336k | 3.34M | 
+| .. Sent       | &nbsp;    | 328 | 33.8k | 336k | 1.53M |
+| .. Dropped    | &nbsp;    | 0   | 0     | 0    | 1.81M |
+
+Here I can make a few important observations.
+
+**Baseline**: One worker (thus, one CPU thread) can sustain 14.88Mpps of L2XC when sFlow is turned off, which
+implies that it has a little bit of CPU left over to do other work, if needed.  With IPv4, I can see
+that the throughput is actually CPU limited: 10.89Mpps can be handled by one worker (thus, one CPU thread). I
+know that MPLS is a little bit more expensive computationally than IPv4, and that checks out. The
+total capacity is 10.11Mpps for one worker, when sFlow is turned off.
+
+**Overhead**: When I turn on sFlow on the interface, VPP will insert the _sflow-node_ into the
+forwarding graph between `device-input` and `ethernet-input`. It means that the sFlow node will see
+_every single_ packet, and it will have to move all of these into the next node, which costs about
+9.5 CPU cycles per packet.  The regression on L2XC is 3.8% but I have to note that VPP was not CPU
+bound on the L2XC so it used some CPU cycles which were still available, before regressing
+throughput. There is an immediate regression of 9.3% on IPv4 and 5.9% on MPLS, only to shuffle the
+packets through the graph.
+
+**Sampling Cost**: But when then doing higher rates of sampling, the further regression is not _that_
+terrible.  Between 1:1'000'000 and 1:10'000, there's barely a noticeable difference. Even in the
+worst case of 1:100, the regression is from 14.32Mpps to 14.15Mpps for L2XC, only 1.2%.  The
+regression for L2XC, IPv4 and MPLS are all very modest, at 1.2% (L2XC), 1.6% (IPv4) and 0.8% (MPLS).
+Of course, by using multiple hardware receive queues and multiple RX workers per interface, the cost
+can be kept well in hand.
+
+**Overload Protection**: At 1:1'000 and an effective rate of 33.65Mpps across all ports, I correctly
+observe 336k samples taken, and sent to PSAMPLE.  At 1:100 however, there are 3.34M samples, but
+they are not fitting through the FIFO, so the plugin is dropping samples to protect downstream
+`sflow-main` thread and `hsflowd`. I can see that here, 1.81M samples have been dropped, while 1.53M
+samples made it through.  By the way, this means VPP is happily sending a whopping 153K samples/sec
+to the collector!
+
+## What's Next
+
+Now that I've seen the UDP packets from our agent to a collector on the wire, and also how
+incredibly efficient the sFlow sampling implementation turned out, I'm super motivated to
+continue the journey with higher level collector receivers like ntopng, sflow-rt or Akvorado. In an
+upcoming article, I'll describe how I rolled out Akvorado at IPng, and what types of changes would
+make the user experience even better (or simpler to understand, at least).
+
+### Acknowledgements
+
+I'd like to thank Neil McKee from inMon for his dedication to getting things right, including the
+finer details such as logging, error handling, API specifications, and documentation. He has been a
+true pleasure to work with and learn from. Also, thank you to the VPP developer community, notably
+Benoit, Florin, Damjan, Dave and Matt, for helping with the review and getting this thing merged in
+time for the 25.02 release.

content/articles/2025-04-09-frysix-evpn.md

@@ -0,0 +1,793 @@
+---
+date: "2025-04-09T07:51:23Z"
+title: 'FrysIX eVPN: think different'
+---
+
+{{< image float="right" src="/assets/frys-ix/frysix-logo-small.png" alt="FrysIX Logo" width="12em" >}}
+
+# Introduction
+
+Somewhere in the far north of the Netherlands, the country where I was born, a town called Jubbega
+is the home of the Frysian Internet Exchange called [[Frys-IX](https://frys-ix.net/)]. Back in 2021,
+a buddy of mine, Arend, said that he was planning on renting a rack at the NIKHEF facility, one of
+the most densely populated facilities in western Europe. He was looking for a few launching
+customers and I was definitely in the market for a presence in Amsterdam. I even wrote about it on
+my [[bucketlist]({{< ref 2021-07-26-bucketlist.md >}})]. Arend and his IT company
+[[ERITAP](https://www.eritap.com/)], took delivery of that rack in May of 2021, and this is when the
+internet exchange with _Frysian roots_ was born. 
+
+In the years from 2021 until now, Arend and I have been operating the exchange with reasonable
+success. It grew from a handful of folks in that first rack, to now some 250 participating ISPs
+with about ten switches in six datacenters across the Amsterdam metro area. It's shifting a cool
+800Gbit of traffic or so. It's dope, and very rewarding to be able to contribute to this community!
+
+## Frys-IX is growing
+
+We have several members with a 2x100G LAG and even though all inter-datacenter links are either dark
+fiber or WDM, we're starting to feel the growing pains as we set our sights to the next step growth.
+You see, when FrysIX did 13.37Gbit of traffic, Arend organized a barbecue. When it did 133.7Gbit of
+traffic, Arend organized an even bigger barbecue. Obviously, the next step is 1337Gbit and joining
+the infamous [[One TeraBit Club](https://github.com/tking/OneTeraBitClub)]. Thomas: we're on our
+way!
+
+It became clear that we will not be able to keep a dependable peering platform if FrysIX remains a
+single L2 broadcast domain, and it also became clear that concatenating multiple 100G ports would be
+operationally expensive (think of all the dark fiber or WDM waves!), and brittle (think of LACP and
+balancing traffic over those ports). We need to modernize in order to stay ahead of the growth
+curve.
+
+## Hello Nokia
+
+{{< image float="right" src="/assets/frys-ix/nokia-7220-d4.png" alt="Nokia 7220-D4" width="20em" >}}
+
+The Nokia 7220 Interconnect Router (7220 IXR) for data center fabric provides fixed-configuration,
+high-capacity platforms that let you bring unmatched scale, flexibility and operational simplicity
+to your data center networks and peering network environments. These devices are built around the
+Broadcom _Trident_ chipset, in the case of the "D4" platform, this is a Trident4 with 28x100G and
+8x400G ports. Whoot!
+
+{{< image float="right" src="/assets/frys-ix/IXR-7220-D3.jpg" alt="Nokia 7220-D3" width="20em" >}}
+
+What I find particularly awesome of the Trident series is their speed (total bandwidth of
+12.8Tbps _per router_), low power use (without optics, the IXR-7220-D4 consumes about 150W) and
+a plethora of advanced capabilities like L2/L3 filtering, IPv4, IPv6 and MPLS routing, and modern
+approaches to scale-out networking such as VXLAN based EVPN. At the FrysIX barbecue in September of
+2024, FrysIX was gifted a rather powerful IXR-7220-D3 router, shown in the picture to the right.
+That's a 32x100G router.
+
+ERITAP has bought two (new in box) IXR-7220-D4 (8x400G,28x100G) routers, and has also acquired two
+IXR-7220-D2 (48x25G,8x100G) routers. So in total, FrysIX is now the proud owner of five of these
+beautiful Nokia devices. If you haven't yet, you should definitely read about these versatile
+routers on the [[Nokia](https://onestore.nokia.com/asset/207599)] website, and some details of the
+_merchant silicon_ switch chips in use on the
+[[Broadcom](https://www.broadcom.com/products/ethernet-connectivity/switching/strataxgs/bcm56880-series)]
+website. 
+
+### eVPN: A small rant
+
+{{< image float="right" src="/assets/frys-ix/FrysIX_ Topology (concept).svg" alt="Topology Concept" width="50%" >}}
+
+First, I need to get something off my chest. Consider a topology for an internet exchange platform,
+taking into account the available equipment, rackspace, power, and cross connects. Somehow, almost
+every design or reference architecture I can find on the Internet, assumes folks want to build a
+[[Clos network](https://en.wikipedia.org/wiki/Clos_network)], which has a topology existing of leaf
+and spine switches. The _spine_ switches have a different set of features than the _leaf_ ones,
+notably they don't have to do provider edge functionality like VXLAN encap and decapsulation.
+Almost all of these designs are showing how one might build a leaf-spine network for hyperscale.
+
+**Critique 1**: my 'spine' (IXR-7220-D4 routers) must also be provider edge.  Practically speaking,
+in the picture above I have these beautiful Nokia IXR-7220-D4 routers, using two 400G ports to
+connect between the facilities, and six 100G ports to connect the smaller breakout switches. That
+would leave a _massive_ amount of capacity unused: 22x 100G and 6x400G ports, to be exact.
+
+**Critique 2**: all 'leaf' (either IXR-7220-D2 routers or Arista switches) can't realistically
+connect to both 'spines'. Our devices are spread out over two (and in practice, more like six)
+datacenters, and it's prohibitively expensive to get 100G waves or dark fiber to create a full mesh.
+It's much more economical to create a star-topology that minimizes cross-datacenter fiber spans.
+
+**Critique 3**: Most of these 'spine-leaf' reference architectures assume that the interior gateway
+protocol is eBGP in what they call the _underlay_, and on top of that, some secondary eBGP that's
+called the _overlay_. Frankly, such a design makes my head spin a little bit. These designs assume
+hundreds of switches, in which case making use of one AS number per switch could make sense, as iBGP
+needs either a 'full mesh', or external route reflectors. 
+
+**Critique 4**: These reference designs also make an assumption that all fiber is local and while
+optics and links can fail, it will be relatively rare to _drain_ a link. However, in
+cross-datacenter networks, draining links for maintenance is very common, for example if the dark
+fiber provider needs to perform repairs on a span that was damaged. With these eBGP-over-eBGP
+connections, traffic engineering is more difficult than simply raising the OSPF (or IS-IS) cost of a
+link, to reroute traffic.
+
+Setting aside eVPN for a second, if I were to build an IP transport network, like I did when I built
+[[IPng Site Local]({{< ref 2023-03-11-mpls-core.md >}})], I would use a much more intuitive
+and simple (I would even dare say elegant) design:
+
+1. Take a classic IGP like [[OSPF](https://en.wikipedia.org/wiki/Open_Shortest_Path_First)], or
+perhaps [[IS-IS](https://en.wikipedia.org/wiki/IS-IS)]. There is no benefit, to me at least, to use
+BGP as an IGP.
+1. I would give each of the links between the switches an IPv4 /31 and enable link-local, and give
+each switch a loopback address with a /32 IPv4 and a /128 IPv6.
+1. If I had multiple links between two given switches, I would probably just use ECMP if my devices
+supported it, and fall back to a LACP signaled bundle-ethernet otherwise.
+1. If I were to need to use BGP (and for eVPN, this need exists), taking the ISP mindset (as opposed
+to the datacenter fabric mindset), I would simply install iBGP against two or three route
+reflectors, and exchange routing information within the same single AS number.
+
+### eVPN: A demo topology
+
+{{< image float="right" src="/assets/frys-ix/Nokia Arista VXLAN.svg" alt="Demo topology" width="50%" >}}
+
+So, that's exactly how I'm going to approach the FrysIX eVPN design: OSPF for the underlay and iBGP
+for the overlay! I have a feeling that some folks will despise me for being contrarian, but you can
+leave your comments below, and don't forget to like-and-subscribe :-)
+
+Arend builds this topology for me in Jubbega - also known as FrysIX HQ. He takes the two
+400G-capable routers and connects them. Then he takes an Arista DCS-7060CX switch, which is eVPN
+capable, with 32x100G ports, based on the Broadcom Tomahawk chipset, and a smaller Nokia
+IXR-7220-D2 with 48x25G and 8x100G ports, based on the Trident3 chipset. He wires all of this up
+to look like the picture on the right.
+
+#### Underlay: Nokia's SR Linux
+
+We boot up the equipment, verify that all the optics and links are up, and connect the management
+ports to an OOB network that I can remotely log in to. This is the first time that either of us work
+on Nokia, but I find it reasonably intuitive once I get a few tips and tricks from Niek. 
+
+```
+[pim@nikhef ~]$ sr_cli
+--{ running }--[  ]--
+A:pim@nikhef# enter candidate
+--{ candidate shared default }--[  ]--
+A:pim@nikhef# set / interface lo0 admin-state enable
+A:pim@nikhef# set / interface lo0 subinterface 0 admin-state enable
+A:pim@nikhef# set / interface lo0 subinterface 0 ipv4 admin-state enable
+A:pim@nikhef# set / interface lo0 subinterface 0 ipv4 address 198.19.16.1/32
+A:pim@nikhef# commit stay
+```
+
+There, my first config snippet! This creates a _loopback_ interface, and similar to JunOS, a
+_subinterface_ (which Juniper calls a _unit_) which enables IPv4 and gives it an /32 address. In SR
+Linux, any interface has to be associated with a _network-instance_, think of those as routing
+domains or VRFs. There's a conveniently named _default_ network-instance, which I'll add this and
+the point-to-point interface between the two 400G routers to:
+
+```
+A:pim@nikhef# info flat interface ethernet-1/29
+set / interface ethernet-1/29 admin-state enable
+set / interface ethernet-1/29 subinterface 0 admin-state enable
+set / interface ethernet-1/29 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/29 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/29 subinterface 0 ipv4 address 198.19.17.1/31
+set / interface ethernet-1/29 subinterface 0 ipv6 admin-state enable
+
+A:pim@nikhef# set / network-instance default type default
+A:pim@nikhef# set / network-instance default admin-state enable
+A:pim@nikhef# set / network-instance default interface ethernet-1/29.0
+A:pim@nikhef# set / network-instance default interface lo0.0
+A:pim@nikhef# commit stay
+```
+
+Cool. Assuming I now also do this on the other IXR-7220-D4 router, called _equinix_ (which gets the
+loopback address 198.19.16.0/32 and the point-to-point on the 400G interface of 198.19.17.0/31), I
+should be able to do my first jumboframe ping:
+
+```
+A:pim@equinix# ping network-instance default 198.19.17.1 -s 9162 -M do
+Using network instance default
+PING 198.19.17.1 (198.19.17.1) 9162(9190) bytes of data.
+9170 bytes from 198.19.17.1: icmp_seq=1 ttl=64 time=0.466 ms
+9170 bytes from 198.19.17.1: icmp_seq=2 ttl=64 time=0.477 ms
+9170 bytes from 198.19.17.1: icmp_seq=3 ttl=64 time=0.547 ms
+```
+
+#### Underlay: SR Linux OSPF
+
+OK, let's get these two Nokia routers to speak OSPF, so that they can reach each other's loopback.
+It's really easy:
+
+```
+A:pim@nikhef# / network-instance default protocols ospf instance default
+--{ candidate shared default }--[ network-instance default protocols ospf instance default ]--
+A:pim@nikhef# set admin-state enable
+A:pim@nikhef# set version ospf-v2
+A:pim@nikhef# set router-id 198.19.16.1
+A:pim@nikhef# set area 0.0.0.0 interface ethernet-1/29.0 interface-type point-to-point
+A:pim@nikhef# set area 0.0.0.0 interface lo0.0 passive true
+A:pim@nikhef# commit stay
+```
+
+Similar to in JunOS, I can descend into a configuration scope: the first line goes into the
+_network-instance_ called `default` and then the _protocols_ called `ospf`, and then the _instance_
+called `default`. Subsequent `set` commands operate at this scope. Once I commit this configuration
+(on the _nikhef_ router and also the _equinix_ router, with its own unique router-id), OSPF quickly
+shoots in action:
+
+```
+A:pim@nikhef# show network-instance default protocols ospf neighbor 
+=========================================================================================
+Net-Inst default OSPFv2 Instance default Neighbors
+=========================================================================================
++---------------------------------------------------------------------------------------+
+| Interface-Name         Rtr Id            State        Pri   RetxQ    Time Before Dead |
++=======================================================================================+
+| ethernet-1/29.0        198.19.16.0       full         1     0        36               |
++---------------------------------------------------------------------------------------+
+-----------------------------------------------------------------------------------------
+No. of Neighbors: 1
+=========================================================================================
+
+A:pim@nikhef# show network-instance default route-table all | more
+IPv4 unicast route table of network instance default
++------------------+-----+------------+--------------+--------+----------+--------+------+-------------+-----------------+
+|        Prefix    | ID  | Route Type |  Route Owner | Active |  Origin  | Metric | Pref | Next-hop    |    Next-hop     |
+|                  |     |            |              |        | Network  |        |      | (Type)      |   Interface     |
+|                  |     |            |              |        | Instance |        |      |             |                 |
++==================+=====+============+==============+========+==========+========+======+=============+=================+
+| 198.19.16.0/32   | 0   | ospfv2     | ospf_mgr     | True   | default  | 1      | 10   | 198.19.17.0 | ethernet-1/29.0 |
+|                  |     |            |              |        |          |        |      | (direct)    |                 |
+| 198.19.16.1/32   | 7   | host       | net_inst_mgr | True   | default  | 0      | 0    | None        | None            |
+| 198.19.17.0/31   | 6   | local      | net_inst_mgr | True   | default  | 0      | 0    | 198.19.17.1 | ethernet-1/29.0 |
+|                  |     |            |              |        |          |        |      | (direct)    |                 |
+| 198.19.17.1/32   | 6   | host       | net_inst_mgr | True   | default  | 0      | 0    | None        | None            |
++==================+=====+============+==============+========+==========+========+======+=============+=================+
+
+A:pim@nikhef# ping network-instance default 198.19.16.0
+Using network instance default
+PING 198.19.16.0 (198.19.16.0) 56(84) bytes of data.
+64 bytes from 198.19.16.0: icmp_seq=1 ttl=64 time=0.484 ms
+64 bytes from 198.19.16.0: icmp_seq=2 ttl=64 time=0.663 ms
+```
+
+Delicious! OSPF has learned the loopback, and it is now reachable. As with most things, going from 0
+to 1 (in this case: understanding how SR Linux works at all) is the most difficult part. Then going
+from 1 to 2 is critical (in this case: making two routers interact with OSPF), but from there on,
+going from 2 to N is easy (in my case: enabling several other point-to-point /31 transit networks on
+the _nikhef_ router, using `ethernet-1/1.0` through `ethernet-1/4.0` with the correct MTU and
+turning on OSPF for these), makes the whole network shoot to life. Slick!
+
+#### Underlay: Arista
+
+I'll point out that one of the devices in this topology is an Arista. We have several of these ready
+for deployment at FrysIX. They are a lot more affordable and easy to find on the second hand /
+refurbished market. These switches come with 32x100G ports, and are really good at packet slinging
+because they're based on the Broadcom _Tomahawk_ chipset. They pack a few less features than the
+_Trident_ chipset that powers the Nokia, but they happen to have all the features we need to run our
+internet exchange .  So I turn my attention to the Arista in the topology. I am much more
+comfortable configuring the whole thing here, as it's not my first time touching these devices:
+
+```
+arista-leaf#show run int loop0
+interface Loopback0
+   ip address 198.19.16.2/32
+   ip ospf area 0.0.0.0
+arista-leaf#show run int Ethernet32/1
+interface Ethernet32/1
+   description Core: Connected to nikhef:ethernet-1/2
+   load-interval 1
+   mtu 9190
+   no switchport
+   ip address 198.19.17.5/31
+   ip ospf cost 1000
+   ip ospf network point-to-point
+   ip ospf area 0.0.0.0
+arista-leaf#show run section router ospf
+router ospf 65500
+   router-id 198.19.16.2
+   redistribute connected
+   network 198.19.0.0/16 area 0.0.0.0
+   max-lsa 12000
+```
+
+I complete the configuration for the other two interfaces on this Arista, port Eth31/1 connects also
+to the _nikhef_ IXR-7220-D4 and I give it a high cost of 1000, while Eth30/1 connects only 1x100G to
+the _nokia-leaf_ IXR-7220-D2 with a cost of 10.
+It's nice to see that OSPF in action - there are two equal path (but high cost) OSPF paths via
+router-id 198.19.16.1 (nikhef), and there's one lower cost path via router-id 198.19.16.3
+(_nokia-leaf_). The traceroute nicely shows the scenic route (arista-leaf -> nokia-leaf -> nokia ->
+equinix). Dope!
+
+```
+arista-leaf#show ip ospf nei
+Neighbor ID     Instance VRF      Pri State                  Dead Time   Address         Interface
+198.19.16.1     65500    default  1   FULL                   00:00:36    198.19.17.4     Ethernet32/1
+198.19.16.3     65500    default  1   FULL                   00:00:31    198.19.17.11    Ethernet30/1
+198.19.16.1     65500    default  1   FULL                   00:00:35    198.19.17.2     Ethernet31/1
+
+arista-leaf#traceroute 198.19.16.0
+traceroute to 198.19.16.0 (198.19.16.0), 30 hops max, 60 byte packets
+ 1  198.19.17.11 (198.19.17.11)  0.220 ms  0.150 ms  0.206 ms
+ 2  198.19.17.6 (198.19.17.6)  0.169 ms  0.107 ms  0.099 ms
+ 3  198.19.16.0 (198.19.16.0)  0.434 ms  0.346 ms  0.303 ms
+```
+
+So far, so good! The _underlay_ is up, every router can reach every other router on its loopback,
+and all OSPF adjacencies are formed. I'll leave the 2x100G between _nikhef_ and _arista-leaf_ at
+high cost for now.
+
+#### Overlay EVPN: SR Linux
+
+The big-picture idea here is to use iBGP with the same private AS number, and because there are two
+main facilities (NIKHEF and Equinix), make each of those bigger IXR-7220-D4 routers act as
+route-reflectors for others. It means that they will have an iBGP session amongst themselves
+(198.191.16.0 <-> 198.19.16.1) and otherwise accept iBGP sessions from any IP address in the
+198.19.16.0/24 subnet.  This way, I don't have to configure any more than strictly necessary on the
+core routers. Any new router can just plug in, form an OSPF adjacency, and connect to both core
+routers. I proceed to configure BGP on the Nokia's like this:
+
+```
+A:pim@nikhef# / network-instance default protocols bgp 
+A:pim@nikhef# set admin-state enable
+A:pim@nikhef# set autonomous-system 65500
+A:pim@nikhef# set router-id 198.19.16.1
+A:pim@nikhef# set dynamic-neighbors accept match 198.19.16.0/24 peer-group overlay
+A:pim@nikhef# set afi-safi evpn admin-state enable
+A:pim@nikhef# set preference ibgp 170
+A:pim@nikhef# set route-advertisement rapid-withdrawal true
+A:pim@nikhef# set route-advertisement wait-for-fib-install false
+A:pim@nikhef# set group overlay peer-as 65500
+A:pim@nikhef# set group overlay afi-safi evpn admin-state enable
+A:pim@nikhef# set group overlay afi-safi ipv4-unicast admin-state disable
+A:pim@nikhef# set group overlay afi-safi ipv6-unicast admin-state disable
+A:pim@nikhef# set group overlay local-as as-number 65500
+A:pim@nikhef# set group overlay route-reflector client true
+A:pim@nikhef# set group overlay transport local-address 198.19.16.1
+A:pim@nikhef# set neighbor 198.19.16.0 admin-state enable
+A:pim@nikhef# set neighbor 198.19.16.0 peer-group overlay
+A:pim@nikhef# commit stay
+```
+
+I can see that iBGP sessions establish between all the devices:
+
+```
+A:pim@nikhef# show network-instance default protocols bgp neighbor
+---------------------------------------------------------------------------------------------------------------------------
+BGP neighbor summary for network-instance "default"              
+Flags: S static, D dynamic, L discovered by LLDP, B BFD enabled, - disabled, * slow
+---------------------------------------------------------------------------------------------------------------------------
+---------------------------------------------------------------------------------------------------------------------------
++-------------+-------------+----------+-------+----------+-------------+---------------+------------+--------------------+
+|   Net-Inst  |    Peer     |  Group   | Flags |  Peer-AS |   State     |      Uptime   |  AFI/SAFI  |   [Rx/Active/Tx]   |
++=============+=============+==========+=======+==========+=============+===============+============+====================+
+| default     | 198.19.16.0 | overlay  | S     | 65500    | established | 0d:0h:2m:32s  | evpn       | [0/0/0]            |
+| default     | 198.19.16.2 | overlay  | D     | 65500    | established | 0d:0h:2m:27s  | evpn       | [0/0/0]            |
+| default     | 198.19.16.3 | overlay  | D     | 65500    | established | 0d:0h:2m:41s  | evpn       | [0/0/0]            |
++-------------+-------------+----------+-------+----------+-------------+---------------+------------+--------------------+
+---------------------------------------------------------------------------------------------------------------------------
+Summary:
+1 configured neighbors, 1 configured sessions are established, 0 disabled peers
+2 dynamic peers
+```
+
+A few things to note here - there one _configured_ neighbor (this is the other IXR-7220-D4 router),
+and two _dynamic_ peers, these are the Arista and the smaller IXR-7220-D2 router. The only address
+family that they are exchanging information for is the _evpn_ family, and no prefixes have been
+learned or sent yet, shown by the `[0/0/0]` designation in the last column.
+
+#### Overlay EVPN: Arista
+
+The Arista is also remarkably straight forward to configure. Here, I'll simply enable the iBGP
+session as follows:
+
+```
+arista-leaf#show run section bgp
+router bgp 65500
+   neighbor evpn peer group
+   neighbor evpn remote-as 65500
+   neighbor evpn update-source Loopback0
+   neighbor evpn ebgp-multihop 3
+   neighbor evpn send-community extended
+   neighbor evpn maximum-routes 12000 warning-only
+   neighbor 198.19.16.0 peer group evpn
+   neighbor 198.19.16.1 peer group evpn
+   !
+   address-family evpn
+      neighbor evpn activate
+
+arista-leaf#show bgp summary 
+BGP summary information for VRF default
+Router identifier 198.19.16.2, local AS number 65500
+Neighbor             AS Session State AFI/SAFI                AFI/SAFI State   NLRI Rcd   NLRI Acc
+----------- ----------- ------------- ----------------------- -------------- ---------- ----------
+198.19.16.0       65500 Established   IPv4 Unicast            Advertised              0          0
+198.19.16.0       65500 Established   L2VPN EVPN              Negotiated              0          0
+198.19.16.1       65500 Established   IPv4 Unicast            Advertised              0          0
+198.19.16.1       65500 Established   L2VPN EVPN              Negotiated              0          0
+```
+
+On this leaf node, I'll have a redundant iBGP session with the two core nodes. Since those core
+nodes are peering amongst themselves, and are configured as route-reflectors, this is all I need. No
+matter how many additional Arista (or Nokia) devices I add to the network, all they'll have to do is
+enable OSPF (so they can reach 198.19.16.0 and .1) and turn on iBGP sessions with both core routers.
+Voila!
+
+#### VXLAN EVPN: SR Linux
+
+Nokia documentation informs me that SR Linux uses a special interface called _system0_ to source its
+VXLAN traffic from, and to add this interface to the _default_ network-instance. So it's a matter of
+defining that interface and associate a VXLAN interface with it, like so:
+
+```
+A:pim@nikhef# set / interface system0 admin-state enable
+A:pim@nikhef# set / interface system0 subinterface 0 admin-state enable
+A:pim@nikhef# set / interface system0 subinterface 0 ipv4 admin-state enable
+A:pim@nikhef# set / interface system0 subinterface 0 ipv4 address 198.19.18.1/32
+A:pim@nikhef# set / network-instance default interface system0.0
+A:pim@nikhef# set / tunnel-interface vxlan1 vxlan-interface 2604 type bridged
+A:pim@nikhef# set / tunnel-interface vxlan1 vxlan-interface 2604 ingress vni 2604
+A:pim@nikhef# set / tunnel-interface vxlan1 vxlan-interface 2604 egress source-ip use-system-ipv4-address
+A:pim@nikhef# commit stay
+```
+
+This creates the plumbing for a VXLAN sub-interface called `vxlan1.2604` which will accept/send
+traffic using VNI 2604 (this happens to be the VLAN id we use at FrysIX for our production Peering
+LAN), and it'll use the `system0.0` address to source that traffic from.
+
+The second part is to create what SR Linux calls a MAC-VRF and put some interface(s) in it:
+
+```
+A:pim@nikhef# set / interface ethernet-1/9 admin-state enable
+A:pim@nikhef# set / interface ethernet-1/9 breakout-mode num-breakout-ports 4
+A:pim@nikhef# set / interface ethernet-1/9 breakout-mode breakout-port-speed 10G
+A:pim@nikhef# set / interface ethernet-1/9/3 admin-state enable
+A:pim@nikhef# set / interface ethernet-1/9/3 vlan-tagging true
+A:pim@nikhef# set / interface ethernet-1/9/3 subinterface 0 type bridged
+A:pim@nikhef# set / interface ethernet-1/9/3 subinterface 0 admin-state enable
+A:pim@nikhef# set / interface ethernet-1/9/3 subinterface 0 vlan encap untagged
+
+A:pim@nikhef# / network-instance peeringlan 
+A:pim@nikhef# set type mac-vrf
+A:pim@nikhef# set admin-state enable
+A:pim@nikhef# set interface ethernet-1/9/3.0
+A:pim@nikhef# set vxlan-interface vxlan1.2604
+A:pim@nikhef# set protocols bgp-evpn bgp-instance 1 admin-state enable
+A:pim@nikhef# set protocols bgp-evpn bgp-instance 1 vxlan-interface vxlan1.2604
+A:pim@nikhef# set protocols bgp-evpn bgp-instance 1 evi 2604
+A:pim@nikhef# set protocols bgp-vpn bgp-instance 1 route-distinguisher rd 65500:2604
+A:pim@nikhef# set protocols bgp-vpn bgp-instance 1 route-target export-rt target:65500:2604
+A:pim@nikhef# set protocols bgp-vpn bgp-instance 1 route-target import-rt target:65500:2604
+A:pim@nikhef# commit stay
+```
+
+In the first block here, Arend took what is a 100G port called `ethernet-1/9` and split it into 4x25G
+ports. Arend forced the port speed to 10G because he has taken a 40G-4x10G DAC, and it happens that
+the third lane is plugged into the Debian machine. So on `ethernet-1/9/3` I'll create a
+sub-interface, make it type _bridged_ (which I've also done on `vxlan1.2604`!) and allow any
+untagged traffic to enter it. 
+
+{{< image width="5em" float="left" src="/assets/shared/brain.png" alt="brain" >}}
+
+If you, like me, are used to either VPP or IOS/XR, this type of sub-interface stuff should feel very
+natural to you. I've written about the sub-interfaces logic on Cisco's IOS/XR and VPP approach in a
+previous [[article]({{< ref 2022-02-14-vpp-vlan-gym.md >}})] which my buddy Fred lovingly calls
+_VLAN Gymnastics_ because the ports are just so damn flexible. Worth a read!
+
+The second block creates a new _network-instance_ which I'll name `peeringlan`, and it associates
+the newly created untagged sub-interface `ethernet-1/9/3.0` with the VXLAN interface, and starts a
+protocol for eVPN instructing traffic in and out of this network-instance to use EVI 2604 on the
+VXLAN sub-interface, and signalling of all MAC addresses learned to use the specified 
+route-distinguisher and import/export route-targets. For simplicity I've just used the same for
+each: 65500:2604.
+
+I continue to add an interface to the `peeringlan` _network-instance_ on the other two Nokia
+routers: `ethernet-1/9/3.0` on the _equinix_ router and `ethernet-1/9.0` on the _nokia-leaf_ router.
+Each of these goes to a 10Gbps port on a Debian machine.
+
+#### VXLAN EVPN: Arista
+
+At this point I'm feeling pretty bullish about the whole project. Arista does not make it very
+difficult on me to configure it for L2 EVPN (which is called MAC-VRF here also):
+
+```
+arista-leaf#conf t
+vlan 2604
+   name v-peeringlan
+interface Ethernet9/3
+   speed forced 10000full
+   switchport access vlan 2604
+
+interface Loopback1
+   ip address 198.19.18.2/32
+interface Vxlan1
+   vxlan source-interface Loopback1
+   vxlan udp-port 4789
+   vxlan vlan 2604 vni 2604
+```
+
+After creating VLAN 2604 and making port Eth9/3 an access port in that VLAN, I'll add a VTEP endpoint
+called `Loopback1`, and a VXLAN interface that uses that to source its traffic. Here, I'll associate
+local VLAN 2604 with the `Vxlan1` and its VNI 2604, to match up with how I configured the Nokias
+previously.
+
+Finally, it's a matter of tying these together by announcing the MAC addresses into the EVPN iBGP
+sessions:
+```
+arista-leaf#conf t
+router bgp 65500
+   vlan 2604
+      rd 65500:2604
+      route-target both 65500:2604
+      redistribute learned
+   !
+```
+
+### Results
+
+To validate the configurations, I learn a cool trick from my buddy Andy on the SR Linux discord
+server. In EOS, I can ask it to check for any obvious mistakes in two places:
+
+```
+arista-leaf#show vxlan config-sanity detail
+Category                            Result  Detail                                            
+---------------------------------- -------- --------------------------------------------------
+Local VTEP Configuration Check        OK                                                      
+  Loopback IP Address                 OK                                                      
+  VLAN-VNI Map                        OK                                                      
+  Flood List                          OK                                                      
+  Routing                             OK                                                      
+  VNI VRF ACL                         OK                                                      
+  Decap VRF-VNI Map                   OK                                                      
+  VRF-VNI Dynamic VLAN                OK                                                      
+Remote VTEP Configuration Check       OK                                                      
+  Remote VTEP                         OK                                                      
+Platform Dependent Check              OK                                                      
+  VXLAN Bridging                      OK                                                      
+  VXLAN Routing                       OK    VXLAN Routing not enabled                         
+CVX Configuration Check               OK                                                      
+  CVX Server                          OK    Not in controller client mode                     
+MLAG Configuration Check              OK    Run 'show mlag config-sanity' to verify MLAG config
+  Peer VTEP IP                        OK    MLAG peer is not connected                        
+  MLAG VTEP IP                        OK                                                      
+  Peer VLAN-VNI                       OK                                                      
+  Virtual VTEP IP                     OK                                                      
+  MLAG Inactive State                 OK                                                      
+
+arista-leaf#show bgp evpn sanity detail 
+Category Check                Status Detail
+-------- -------------------- ------ ------
+General  Send community       OK           
+General  Multi-agent mode     OK           
+General  Neighbor established OK           
+L2       MAC-VRF route-target OK           
+         import and export                 
+L2       MAC-VRF              OK           
+         route-distinguisher               
+L2       MAC-VRF redistribute OK           
+L2       MAC-VRF overlapping  OK           
+         VLAN                              
+L2       Suppressed MAC       OK           
+VXLAN    VLAN to VNI map for  OK           
+         MAC-VRF                           
+VXLAN    VRF to VNI map for   OK           
+         IP-VRF                            
+```
+
+#### Results: Arista view
+
+Inspecting the MAC addresses learned from all four of the client ports on the Debian machine is
+easy:
+
+```
+arista-leaf#show bgp evpn summary 
+BGP summary information for VRF default
+Router identifier 198.19.16.2, local AS number 65500
+Neighbor Status Codes: m - Under maintenance
+  Neighbor    V AS           MsgRcvd   MsgSent  InQ OutQ  Up/Down State   PfxRcd PfxAcc
+  198.19.16.0 4 65500           3311      3867    0    0 18:06:28 Estab   7      7
+  198.19.16.1 4 65500           3308      3873    0    0 18:06:28 Estab   7      7
+
+arista-leaf#show bgp evpn vni 2604 next-hop 198.19.18.3
+BGP routing table information for VRF default
+Router identifier 198.19.16.2, local AS number 65500
+Route status codes: * - valid, > - active, S - Stale, E - ECMP head, e - ECMP
+                    c - Contributing to ECMP, % - Pending BGP convergence
+Origin codes: i - IGP, e - EGP, ? - incomplete
+AS Path Attributes: Or-ID - Originator ID, C-LST - Cluster List, LL Nexthop - Link Local Nexthop
+
+          Network                Next Hop              Metric  LocPref Weight  Path
+ * >Ec    RD: 65500:2604 mac-ip e43a.6e5f.0c59
+                                 198.19.18.3           -       100     0       i Or-ID: 198.19.16.3 C-LST: 198.19.16.1 
+ *  ec    RD: 65500:2604 mac-ip e43a.6e5f.0c59
+                                 198.19.18.3           -       100     0       i Or-ID: 198.19.16.3 C-LST: 198.19.16.0 
+ * >Ec    RD: 65500:2604 imet 198.19.18.3
+                                 198.19.18.3           -       100     0       i Or-ID: 198.19.16.3 C-LST: 198.19.16.1 
+ *  ec    RD: 65500:2604 imet 198.19.18.3
+                                 198.19.18.3           -       100     0       i Or-ID: 198.19.16.3 C-LST: 198.19.16.0 
+```
+There's a lot to unpack here! The Arista is seeing that from the _route-distinguisher_ I configured
+on all the sessions, it is learning one MAC address on neighbor 198.19.18.3 (this is the VTEP for
+the _nokia-leaf_ router) from both iBGP sessions. The MAC address is learned from originator
+198.19.16.3 (the loopback of the _nokia-leaf_ router), from two cluster members, the active one on
+iBGP speaker 198.19.16.1 (_nikhef_) and a backup member on 198.19.16.0 (_equinix_).
+
+I can also see that there's a bunch of `imet` route entries, and Andy explained these to me. They are
+a signal from a VTEP participant that they are interested in seeing multicast traffic (like neighbor
+discovery or ARP requests) flooded to them. Every router participating in this L2VPN will raise such
+an `imet` route, which I'll see in duplicates as well (one from each iBGP session). This checks out.
+
+#### Results: SR Linux view
+
+The Nokia IXR-7220-D4 router called _equinix_ has also learned a bunch of EVPN routing entries,
+which I can inspect as follows:
+
+```
+A:pim@equinix# show network-instance default protocols bgp routes evpn route-type summary 
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------
+Show report for the BGP route table of network-instance "default"
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------
+Status codes: u=used, *=valid, >=best, x=stale, b=backup
+Origin codes: i=IGP, e=EGP, ?=incomplete
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------
+BGP Router ID: 198.19.16.0      AS: 65500      Local AS: 65500
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------
+Type 2 MAC-IP Advertisement Routes
++--------+---------------+--------+-------------------+------------+-------------+------+-------------+--------+--------------------------------+------------------+
+| Status |    Route-     | Tag-ID |    MAC-address    | IP-address | neighbor    | Path-|  Next-Hop   |  Label |              ESI               |   MAC Mobility   |
+|        | distinguisher |        |                   |            |             |   id |             |        |                                |                  |
++========+===============+========+===================+============+=============+======+============-+========+================================+==================+
+| u*>    | 65500:2604    | 0      | E4:3A:6E:5F:0C:57 | 0.0.0.0    | 198.19.16.1 | 0    | 198.19.18.1 | 2604   | 00:00:00:00:00:00:00:00:00:00  | -                |
+| *      | 65500:2604    | 0      | E4:3A:6E:5F:0C:58 | 0.0.0.0    | 198.19.16.1 | 0    | 198.19.18.2 | 2604   | 00:00:00:00:00:00:00:00:00:00  | -                |
+| u*>    | 65500:2604    | 0      | E4:3A:6E:5F:0C:58 | 0.0.0.0    | 198.19.16.2 | 0    | 198.19.18.2 | 2604   | 00:00:00:00:00:00:00:00:00:00  | -                |
+| *      | 65500:2604    | 0      | E4:3A:6E:5F:0C:59 | 0.0.0.0    | 198.19.16.1 | 0    | 198.19.18.3 | 2604   | 00:00:00:00:00:00:00:00:00:00  | -                |
+| u*>    | 65500:2604    | 0      | E4:3A:6E:5F:0C:59 | 0.0.0.0    | 198.19.16.3 | 0    | 198.19.18.3 | 2604   | 00:00:00:00:00:00:00:00:00:00  | -                |
++--------+---------------+--------+-------------------+------------+-------------+------+-------------+--------+--------------------------------+------------------+
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------
+Type 3 Inclusive Multicast Ethernet Tag Routes
++--------+-----------------------------+--------+---------------------+-----------------+--------+-----------------------+
+| Status |     Route-distinguisher     | Tag-ID |    Originator-IP    |    neighbor     | Path-  |       Next-Hop        |
+|        |                             |        |                     |                 |   id   |                       |
++========+=============================+========+=====================+=================+========+=======================+
+| u*>    | 65500:2604                  | 0      | 198.19.18.1         | 198.19.16.1     | 0      | 198.19.18.1           |
+| *      | 65500:2604                  | 0      | 198.19.18.2         | 198.19.16.1     | 0      | 198.19.18.2           |
+| u*>    | 65500:2604                  | 0      | 198.19.18.2         | 198.19.16.2     | 0      | 198.19.18.2           |
+| *      | 65500:2604                  | 0      | 198.19.18.3         | 198.19.16.1     | 0      | 198.19.18.3           |
+| u*>    | 65500:2604                  | 0      | 198.19.18.3         | 198.19.16.3     | 0      | 198.19.18.3           |
++--------+-----------------------------+--------+---------------------+-----------------+--------+-----------------------+
+--------------------------------------------------------------------------------------------------------------------------
+0 Ethernet Auto-Discovery routes 0 used, 0 valid
+5 MAC-IP Advertisement routes 3 used, 5 valid
+5 Inclusive Multicast Ethernet Tag routes 3 used, 5 valid
+0 Ethernet Segment routes 0 used, 0 valid
+0 IP Prefix routes 0 used, 0 valid
+0 Selective Multicast Ethernet Tag routes 0 used, 0 valid
+0 Selective Multicast Membership Report Sync routes 0 used, 0 valid
+0 Selective Multicast Leave Sync routes 0 used, 0 valid
+--------------------------------------------------------------------------------------------------------------------------
+```
+
+I have to say, SR Linux output is incredibly verbose! But, I can see all the relevant bits and bobs
+here.  Each MAC-IP entry is accounted for, I can see several nexthops pointing at the nikhef switch,
+one pointing at the nokia-leaf router and one pointing at the Arista switch. I also see the `imet`
+entries. One thing to note -- the SR Linux implementation leaves the type-2 routes empty with a
+0.0.0.0 IPv4 address, while the Arista (in my opinion, more correctly) leaves them as NULL
+(unspecified). But, everything looks great!
+
+#### Results: Debian view
+
+There's one more thing to show, and that's kind of the 'proof is in the pudding' moment. As I said,
+Arend hooked up a Debian machine with an Intel X710-DA4 network card, which sports 4x10G SFP+
+connections.  This network card is a regular in my AS8298 network, as it has excellent DPDK support
+and can easily pump 40Mpps with VPP. IPng 🥰 Intel X710!
+
+```
+root@debian:~ # ip netns add nikhef
+root@debian:~ # ip link set enp1s0f0 netns nikhef
+root@debian:~ # ip netns exec nikhef ip link set enp1s0f0 up mtu 9000
+root@debian:~ # ip netns exec nikhef ip addr add 192.0.2.10/24 dev enp1s0f0
+root@debian:~ # ip netns exec nikhef ip addr add 2001:db8::10/64 dev enp1s0f0
+
+root@debian:~ # ip netns add arista-leaf
+root@debian:~ # ip link set enp1s0f1 netns arista-leaf
+root@debian:~ # ip netns exec arista-leaf ip link set enp1s0f1 up mtu 9000
+root@debian:~ # ip netns exec arista-leaf ip addr add 192.0.2.11/24 dev enp1s0f1
+root@debian:~ # ip netns exec arista-leaf ip addr add 2001:db8::11/64 dev enp1s0f1
+
+root@debian:~ # ip netns add nokia-leaf
+root@debian:~ # ip link set enp1s0f2 netns nokia-leaf
+root@debian:~ # ip netns exec nokia-leaf ip link set enp1s0f2 up mtu 9000
+root@debian:~ # ip netns exec nokia-leaf ip addr add 192.0.2.12/24 dev enp1s0f2
+root@debian:~ # ip netns exec nokia-leaf ip addr add 2001:db8::12/64 dev enp1s0f2
+
+root@debian:~ # ip netns add equinix
+root@debian:~ # ip link set enp1s0f3 netns equinix
+root@debian:~ # ip netns exec equinix ip link set enp1s0f3 up mtu 9000
+root@debian:~ # ip netns exec equinix ip addr add 192.0.2.13/24 dev enp1s0f3 
+root@debian:~ # ip netns exec equinix ip addr add 2001:db8::13/64 dev enp1s0f3 
+
+root@debian:~# ip netns exec nikhef fping -g 192.0.2.8/29
+192.0.2.10 is alive
+192.0.2.11 is alive
+192.0.2.12 is alive
+192.0.2.13 is alive
+
+root@debian:~# ip netns exec arista-leaf fping 2001:db8::10 2001:db8::11 2001:db8::12 2001:db8::13
+2001:db8::10 is alive
+2001:db8::11 is alive
+2001:db8::12 is alive
+2001:db8::13 is alive
+
+root@debian:~# ip netns exec equinix ip nei
+192.0.2.10 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:57 STALE 
+192.0.2.11 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:58 STALE 
+192.0.2.12 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:59 STALE 
+fe80::e63a:6eff:fe5f:c57 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:57 STALE 
+fe80::e63a:6eff:fe5f:c58 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:58 STALE 
+fe80::e63a:6eff:fe5f:c59 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:59 STALE 
+2001:db8::10 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:57 STALE 
+2001:db8::11 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:58 STALE 
+2001:db8::12 dev enp1s0f3 lladdr e4:3a:6e:5f:0c:59 STALE 
+```
+
+The Debian machine puts each network card into its own network namespace, and gives them both an IPv4
+and an IPv6 address. I can then enter the `nikhef` network namespace, which has its NIC connected to
+the IXR-7220-D4 router called _nikhef_, and ping all four endpoints. Similarly, I can enter the
+`arista-leaf` namespace and ping6 all four endpoints. Finally, I take a look at the IPv6 and IPv4
+neighbor table on the network card that is connected to the _equinix_ router. All three MAC addresses are
+seen. This proves end to end connectivity across the EVPN VXLAN, and full interoperability. Booyah!
+
+Performance? We got that! I'm not worried as these Nokia routers are rated for 12.8Tbps of VXLAN....
+```
+root@debian:~# ip netns exec equinix iperf3 -c 192.0.2.12 
+Connecting to host 192.0.2.12, port 5201
+[  5] local 192.0.2.10 port 34598 connected to 192.0.2.12 port 5201
+[ ID] Interval           Transfer     Bitrate         Retr  Cwnd
+[  5]   0.00-1.00   sec  1.15 GBytes  9.91 Gbits/sec   19   1.52 MBytes       
+[  5]   1.00-2.00   sec  1.15 GBytes  9.90 Gbits/sec    3   1.54 MBytes       
+[  5]   2.00-3.00   sec  1.15 GBytes  9.90 Gbits/sec    1   1.54 MBytes       
+[  5]   3.00-4.00   sec  1.15 GBytes  9.90 Gbits/sec    1   1.54 MBytes       
+[  5]   4.00-5.00   sec  1.15 GBytes  9.90 Gbits/sec    0   1.54 MBytes       
+[  5]   5.00-6.00   sec  1.15 GBytes  9.90 Gbits/sec    0   1.54 MBytes       
+[  5]   6.00-7.00   sec  1.15 GBytes  9.90 Gbits/sec    0   1.54 MBytes       
+[  5]   7.00-8.00   sec  1.15 GBytes  9.90 Gbits/sec    0   1.54 MBytes       
+[  5]   8.00-9.00   sec  1.15 GBytes  9.90 Gbits/sec    0   1.54 MBytes       
+[  5]   9.00-10.00  sec  1.15 GBytes  9.90 Gbits/sec    0   1.54 MBytes       
+- - - - - - - - - - - - - - - - - - - - - - - - -
+[ ID] Interval           Transfer     Bitrate         Retr
+[  5]   0.00-10.00  sec  11.5 GBytes  9.90 Gbits/sec   24             sender
+[  5]   0.00-10.00  sec  11.5 GBytes  9.90 Gbits/sec                  receiver
+
+iperf Done.
+```
+
+## What's Next
+
+There's a few improvements I can make before deploying this architecture to the internet exchange.
+Notably:
+*   the functional equivalent of _port security_, that is to say only allowing one or two MAC
+    addresses per member port. FrysIX has a strict one-port-one-member-one-MAC rule, and having port
+    security will greatly improve our resilience.
+*   SR Linux has the ability to suppress ARP, _even on L2 MAC-VRF_! It's relatively well known for
+    IRB based setups, but adding this to transparent bridge-domains is possible in Nokia
+[[ref](https://documentation.nokia.com/srlinux/22-6/SR_Linux_Book_Files/EVPN-VXLAN_Guide/services-evpn-vxlan-l2.html#configuring_evpn_learning_for_proxy_arp)],
+    using the syntax of `protocols bgp-evpn bgp-instance 1 routes bridge-table mac-ip advertise
+true`. This will glean the IP addresses based on intercepted ARP requests, and reduce the need for
+    BUM flooding.
+*   Andy informs me that Arista also has this feature. By setting `router l2-vpn` and `arp learning bridged`,
+    the suppression of ARP requests/replies also works in the same way. This greatly reduces cross-router
+    BUM flooding. If DE-CIX can do it, so can FrysIX :)
+*   some automation - although configuring the MAC-VRF across Arista and SR Linux is  definitely not
+    as difficult as I thought, having some automation in place will avoid errors and mistakes. It
+    would suck if the IXP collapsed because I botched a link drain or PNI configuration!
+
+### Acknowledgements
+
+I am relatively new to EVPN configurations, and wanted to give a shoutout to Andy Whitaker who
+jumped in very quickly when I asked a question on the SR Linux Discord. He was gracious with his
+time and spent a few hours on a video call with me, explaining EVPN in great detail both for Arista
+as well as SR Linux, and in particular wanted to give a big "Thank you!" for helping me understand
+symmetric and asymmetric IRB in the context of multivendor EVPN. Andy is about to start a new job at
+Nokia, and I wish him all the best. To my friends at Nokia: you caught a good one, Andy is pure
+gold!
+
+I also want to thank Niek for helping me take my first baby steps onto this platform and patiently
+answering my nerdly questions about the platform, the switch chip, and the configuration philosophy.
+Learning a new NOS is always a fun task, and it was made super fun because Niek spent an hour with
+Arend and I on a video call, giving a bunch of operational tips and tricks along the way.
+
+Finally, Arend and ERITAP are an absolute joy to work with. We took turns hacking on the lab, which
+Arend made available for me while I am traveling to Mississippi this week. Thanks for the kWh and
+OOB access, and for brainstorming the config with me!
+
+### Reference configurations
+
+Here's the configs for all machines in this demonstration:
+[[nikhef](/assets/frys-ix/nikhef.conf)] | [[equinix](/assets/frys-ix/equinix.conf)] | [[nokia-leaf](/assets/frys-ix/nokia-leaf.conf)] | [[arista-leaf](/assets/frys-ix/arista-leaf.conf)] 

content/articles/2025-05-03-containerlab-1.md

@@ -0,0 +1,464 @@
+---
+date: "2025-05-03T15:07:23Z"
+title: 'VPP in Containerlab - Part 1'
+---
+
+{{< image float="right" src="/assets/containerlab/containerlab.svg" alt="Containerlab Logo" width="12em" >}}
+
+# Introduction
+
+From time to time the subject of containerized VPP instances comes up. At IPng, I run the routers in
+AS8298 on bare metal (Supermicro and Dell hardware), as it allows me to maximize performance.
+However, VPP is quite friendly in virtualization. Notably, it runs really well on virtual machines
+like Qemu/KVM or VMWare. I can pass through PCI devices directly to the host, and use CPU pinning to
+allow the guest virtual machine access to the underlying physical hardware. In such a mode, VPP
+performance almost the same as on bare metal. But did you know that VPP can also run in Docker?
+
+The other day I joined the [[ZANOG'25](https://nog.net.za/event1/zanog25/)] in Durban, South Africa.
+One of the presenters was Nardus le Roux of Nokia, and he showed off a project called
+[[Containerlab](https://containerlab.dev/)], which provides a CLI for orchestrating and managing
+container-based networking labs. It starts the containers, builds a virtual wiring between them to
+create lab topologies of users choice and manages labs lifecycle.
+
+Quite regularly I am asked 'when will you add VPP to Containerlab?', but at ZANOG I made a promise
+to actually add them. Here I go, on a journey to integrate VPP into Containerlab!
+
+## Containerized VPP
+
+The folks at [[Tigera](https://www.tigera.io/project-calico/)] maintain a project called _Calico_,
+which accelerates Kubernetes CNI (Container Network Interface) by using [[FD.io](https://fd.io)]
+VPP. Since the origins of Kubernetes are to run containers in a Docker environment, it stands to
+reason that it should be possible to run a containerized VPP. I start by reading up on how they
+create their Docker image, and I learn a lot.
+
+### Docker Build
+
+Considering IPng runs bare metal Debian (currently Bookworm) machines, my Docker image will be based
+on `debian:bookworm` as well. The build starts off quite modest:
+
+```
+pim@summer:~$ mkdir -p src/vpp-containerlab
+pim@summer:~/src/vpp-containerlab$ cat < EOF > Dockerfile.bookworm
+FROM debian:bookworm
+ARG DEBIAN_FRONTEND=noninteractive
+ARG VPP_INSTALL_SKIP_SYSCTL=true
+ARG REPO=release
+RUN apt-get update && apt-get -y install curl procps && apt-get clean
+
+# Install VPP
+RUN curl -s https://packagecloud.io/install/repositories/fdio/${REPO}/script.deb.sh | bash
+RUN apt-get update && apt-get -y install vpp vpp-plugin-core && apt-get clean
+
+CMD ["/usr/bin/vpp","-c","/etc/vpp/startup.conf"]
+EOF
+pim@summer:~/src/vpp-containerlab$ docker build -f Dockerfile.bookworm . -t pimvanpelt/vpp-containerlab
+```
+
+One gotcha - when I install the upstream VPP debian packages, they generate a `sysctl` file which it
+tries to execute. However, I can't set sysctl's in the container, so the build fails. I take a look
+at the VPP source code and find `src/pkg/debian/vpp.postinst` which helpfully contains a means to
+override setting the sysctl's, using an environment variable called `VPP_INSTALL_SKIP_SYSCTL`.
+
+### Running VPP in Docker
+
+With the Docker image built, I need to tweak the VPP startup configuration a little bit, to allow it
+to run well in a Docker environment. There are a few things I make note of:
+1.   We may not have huge pages on the host machine, so I'll set all the page sizes to the
+     linux-default 4kB rather than 2MB or 1GB hugepages. This creates a performance regression, but
+     in the case of Containerlab, we're not here to build high performance stuff, but rather users
+     will be doing functional testing.
+1.   DPDK requires either UIO of VFIO kernel drivers, so that it can bind its so-called _poll mode
+     driver_  to the network cards. It also requires huge pages. Since my first version will be
+     using only virtual ethernet interfaces, I'll disable DPDK and VFIO alltogether.
+1.   VPP can run any number of CPU worker threads. In its simplest form, I can also run it with only
+     one thread. Of course, this will not be a high performance setup, but since I'm already not
+     using hugepages, I'll use only 1 thread.
+
+The VPP `startup.conf` configuration file I came up with:
+
+```
+pim@summer:~/src/vpp-containerlab$ cat < EOF > clab-startup.conf
+unix {
+  interactive
+  log /var/log/vpp/vpp.log
+  full-coredump
+  cli-listen /run/vpp/cli.sock
+  cli-prompt vpp-clab#
+  cli-no-pager
+  poll-sleep-usec 100
+}
+
+api-trace {
+  on
+}
+
+memory {
+  main-heap-size 512M
+  main-heap-page-size 4k
+}
+buffers {
+  buffers-per-numa 16000
+  default data-size 2048
+  page-size 4k
+}
+
+statseg {
+  size 64M
+  page-size 4k
+  per-node-counters on
+}
+
+plugins {
+  plugin default { enable }
+  plugin dpdk_plugin.so { disable }
+}
+EOF
+```
+
+Just a couple of notes for those who are running VPP in production. Each of the `*-page-size` config
+settings take the normal Linux pagesize of 4kB, which effectively avoids VPP from using anhy
+hugepages. Then, I'll specifically disable the DPDK plugin, although I didn't install it in the
+Dockerfile build, as it lives in its own dedicated Debian package called `vpp-plugin-dpdk`. Finally,
+I'll make VPP use less CPU by telling it to sleep for 100 microseconds between each poll iteration.
+In production environments, VPP will use 100% of the CPUs it's assigned, but in this lab, it will
+not be quite as hungry. By the way, even in this sleepy mode, it'll still easily handle a gigabit
+of traffic!
+
+Now, VPP wants to run as root and it needs a few host features, notably tuntap devices and vhost,
+and a few capabilites, notably NET_ADMIN and SYS_PTRACE. I take a look at the
+[[manpage](https://man7.org/linux/man-pages/man7/capabilities.7.html)]:
+*   ***CAP_SYS_NICE***: allows to set real-time scheduling, CPU affinity, I/O scheduling class, and
+    to migrate and move memory pages.
+*   ***CAP_NET_ADMIN***: allows to perform various network-relates operations like interface
+    configs, routing tables, nested network namespaces, multicast, set promiscuous mode, and so on.
+*   ***CAP_SYS_PTRACE***: allows to trace arbitrary processes using `ptrace(2)`, and a few related
+    kernel system calls.
+
+Being a networking dataplane implementation, VPP wants to be able to tinker with network devices.
+This is not typically allowed in Docker containers, although the Docker developers did make some
+consessions for those containers that need just that little bit more access. They described it in
+their
+[[docs](https://docs.docker.com/engine/containers/run/#runtime-privilege-and-linux-capabilities)] as
+follows:
+
+| The --privileged flag gives all capabilities to the container. When the operator executes docker
+| run --privileged, Docker enables access to all devices on the host, and reconfigures AppArmor or
+| SELinux to allow the container nearly all the same access to the host as processes running outside
+| containers on the host. Use this flag with caution. For more information about the --privileged
+| flag, see the docker run reference.
+
+{{< image width="4em" float="left" src="/assets/shared/warning.png" alt="Warning" >}}
+In this moment, I feel I should point out that running a Docker container with `--privileged` flag
+set does give it _a lot_ of privileges. A container with `--privileged` is not a securely sandboxed
+process. Containers in this mode can get a root shell on the host and take control over the system.
+
+With that little fineprint warning out of the way, I am going to Yolo like a boss:
+
+```
+pim@summer:~/src/vpp-containerlab$ docker run --name clab-pim \
+                --cap-add=NET_ADMIN --cap-add=SYS_NICE --cap-add=SYS_PTRACE \
+                --device=/dev/net/tun:/dev/net/tun --device=/dev/vhost-net:/dev/vhost-net \
+                --privileged -v $(pwd)/clab-startup.conf:/etc/vpp/startup.conf:ro \
+                docker.io/pimvanpelt/vpp-containerlab
+clab-pim
+```
+
+### Configuring VPP in Docker
+
+And with that, the Docker container is running! I post a screenshot on
+[[Mastodon](https://ublog.tech/@IPngNetworks/114392852468494211)] and my buddy John responds with a
+polite but firm insistence that I explain myself. Here you go, buddy :)
+
+In another terminal, I can play around with this VPP instance a little bit:
+```
+pim@summer:~$ docker exec -it clab-pim bash
+root@d57c3716eee9:/# ip -br l
+lo               UNKNOWN        00:00:00:00:00:00 <LOOPBACK,UP,LOWER_UP> 
+eth0@if530566    UP             02:42:ac:11:00:02 <BROADCAST,MULTICAST,UP,LOWER_UP> 
+
+root@d57c3716eee9:/# ps auxw   
+USER         PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
+root           1  2.2  0.2 17498852 160300 ?     Rs   15:11   0:00 /usr/bin/vpp -c /etc/vpp/startup.conf
+root          10  0.0  0.0   4192  3388 pts/0    Ss   15:11   0:00 bash
+root          18  0.0  0.0   8104  4056 pts/0    R+   15:12   0:00 ps auxw
+
+root@d57c3716eee9:/# vppctl
+    _______    _        _   _____  ___ 
+ __/ __/ _ \  (_)__    | | / / _ \/ _ \
+ _/ _// // / / / _ \   | |/ / ___/ ___/
+ /_/ /____(_)_/\___/   |___/_/  /_/    
+
+vpp-clab# show version
+vpp v25.02-release built by root on d5cd2c304b7f at 2025-02-26T13:58:32
+vpp-clab# show interfaces
+              Name               Idx    State  MTU (L3/IP4/IP6/MPLS)     Counter          Count     
+local0                            0     down          0/0/0/0       
+```
+
+Slick! I can see that the container has an `eth0` device, which Docker has connected to the main
+bridged network. For now, there's only one process running, pid 1 proudly shows VPP (as in Docker,
+the `CMD` field will simply replace `init`. Later on, I can imagine running a few more daemons like
+SSH and so on, but for now, I'm happy.
+
+Looking at VPP itself, it has no network interfaces yet, except for the default `local0` interface.
+
+### Adding Interfaces in Docker
+
+But if I don't have DPDK, how will I add interfaces? Enter `veth(4)`. From the
+[[manpage](https://man7.org/linux/man-pages/man4/veth.4.html)], I learn that veth devices are
+virtual Ethernet devices.  They can act as tunnels between network namespaces to create a bridge to
+a physical network device in another namespace, but can also be used as standalone network devices.
+veth devices are always created in interconnected pairs.
+
+Of course, Docker users will recognize this. It's like bread and butter for containers to
+communicate with one another - and with the host they're running on. I can simply create a Docker
+network and attach one half of it to a running container, like so:
+
+```
+pim@summer:~$ docker network create --driver=bridge clab-network \
+                     --subnet 192.0.2.0/24 --ipv6 --subnet 2001:db8::/64
+5711b95c6c32ac0ed185a54f39e5af4b499677171ff3d00f99497034e09320d2
+pim@summer:~$ docker network connect clab-network clab-pim --ip '' --ip6 ''
+```
+
+The first command here creates a new network called `clab-network` in Docker. As a result, a new
+bridge called `br-5711b95c6c32` shows up on the host. The bridge name is chosen from the UUID of the
+Docker object. Seeing as I added an IPv4 and IPv6 subnet to the bridge, it gets configured with the
+first address in both:
+
+```
+pim@summer:~/src/vpp-containerlab$ brctl show br-5711b95c6c32
+bridge name       bridge id               STP enabled     interfaces
+br-5711b95c6c32   8000.0242099728c6       no              veth021e363
+
+
+pim@summer:~/src/vpp-containerlab$ ip -br a show dev br-5711b95c6c32
+br-5711b95c6c32  UP     192.0.2.1/24 2001:db8::1/64 fe80::42:9ff:fe97:28c6/64 fe80::1/64 
+```
+
+The second command creates a `veth` pair, and puts one half of it in the bridge, and this interface
+is called `veth021e363` above. The other half of it pops up as `eth1` in the Docker container:
+
+```
+pim@summer:~/src/vpp-containerlab$ docker exec -it clab-pim bash
+root@d57c3716eee9:/# ip -br l
+lo               UNKNOWN        00:00:00:00:00:00 <LOOPBACK,UP,LOWER_UP> 
+eth0@if530566    UP             02:42:ac:11:00:02 <BROADCAST,MULTICAST,UP,LOWER_UP> 
+eth1@if530577    UP             02:42:c0:00:02:02 <BROADCAST,MULTICAST,UP,LOWER_UP> 
+```
+
+One of the many awesome features of VPP is its ability to attach to these `veth` devices by means of
+its `af-packet` driver, by reusing the same MAC address (in this case `02:42:c0:00:02:02`). I first
+take a look at the linux [[manpage](https://man7.org/linux/man-pages/man7/packet.7.html)] for it,
+and then read up on the VPP
+[[documentation](https://fd.io/docs/vpp/v2101/gettingstarted/progressivevpp/interface)] on the
+topic.
+
+
+However, my attention is drawn to Docker assigning an IPv4 and IPv6 address to the container:
+```
+root@d57c3716eee9:/# ip -br a
+lo               UNKNOWN        127.0.0.1/8 ::1/128
+eth0@if530566    UP             172.17.0.2/16
+eth1@if530577    UP             192.0.2.2/24 2001:db8::2/64 fe80::42:c0ff:fe00:202/64
+root@d57c3716eee9:/# ip addr del 192.0.2.2/24  dev eth1
+root@d57c3716eee9:/# ip addr del 2001:db8::2/64 dev eth1
+```
+
+I decide to remove them from here, as in the end, `eth1` will be owned by VPP so _it_ should be
+setting the IPv4 and IPv6 addresses. For the life of me, I don't see how I can avoid Docker from
+assinging IPv4 and IPv6 addresses to this container ... and the
+[[docs](https://docs.docker.com/engine/network/)] seem to be off as well, as they suggest I can pass
+a flagg `--ipv4=False` but that flag doesn't exist, at least not on my Bookworm Docker variant. I
+make a mental note to discuss this with the folks in the Containerlab community.
+
+
+Anyway, armed with this knowledge I can bind the container-side veth pair called `eth1` to VPP, like
+so:
+
+```
+root@d57c3716eee9:/# vppctl
+    _______    _        _   _____  ___ 
+ __/ __/ _ \  (_)__    | | / / _ \/ _ \
+ _/ _// // / / / _ \   | |/ / ___/ ___/
+ /_/ /____(_)_/\___/   |___/_/  /_/    
+
+vpp-clab# create host-interface name eth1 hw-addr 02:42:c0:00:02:02
+vpp-clab# set interface name host-eth1 eth1
+vpp-clab# set interface mtu 1500 eth1
+vpp-clab# set interface ip address eth1 192.0.2.2/24
+vpp-clab# set interface ip address eth1 2001:db8::2/64
+vpp-clab# set interface state eth1 up
+vpp-clab# show int addr
+eth1 (up):
+  L3 192.0.2.2/24
+  L3 2001:db8::2/64
+local0 (dn):
+```
+
+## Results
+
+After all this work, I've successfully created a Docker image based on Debian Bookworm and VPP 25.02
+(the current stable release version), started a container with it, added a network bridge in Docker,
+which binds the host `summer` to the container. Proof, as they say, is in the ping-pudding:
+
+```
+pim@summer:~/src/vpp-containerlab$ ping -c5 2001:db8::2
+PING 2001:db8::2(2001:db8::2) 56 data bytes
+64 bytes from 2001:db8::2: icmp_seq=1 ttl=64 time=0.113 ms
+64 bytes from 2001:db8::2: icmp_seq=2 ttl=64 time=0.056 ms
+64 bytes from 2001:db8::2: icmp_seq=3 ttl=64 time=0.202 ms
+64 bytes from 2001:db8::2: icmp_seq=4 ttl=64 time=0.102 ms
+64 bytes from 2001:db8::2: icmp_seq=5 ttl=64 time=0.100 ms
+
+--- 2001:db8::2 ping statistics ---
+5 packets transmitted, 5 received, 0% packet loss, time 4098ms
+rtt min/avg/max/mdev = 0.056/0.114/0.202/0.047 ms
+pim@summer:~/src/vpp-containerlab$ ping -c5 192.0.2.2
+PING 192.0.2.2 (192.0.2.2) 56(84) bytes of data.
+64 bytes from 192.0.2.2: icmp_seq=1 ttl=64 time=0.043 ms
+64 bytes from 192.0.2.2: icmp_seq=2 ttl=64 time=0.032 ms
+64 bytes from 192.0.2.2: icmp_seq=3 ttl=64 time=0.019 ms
+64 bytes from 192.0.2.2: icmp_seq=4 ttl=64 time=0.041 ms
+64 bytes from 192.0.2.2: icmp_seq=5 ttl=64 time=0.027 ms
+
+--- 192.0.2.2 ping statistics ---
+5 packets transmitted, 5 received, 0% packet loss, time 4063ms
+rtt min/avg/max/mdev = 0.019/0.032/0.043/0.008 ms
+```
+
+And in case that simple ping-test wasn't enough to get you excited, here's a packet trace from VPP
+itself, while I'm performing this ping:
+
+```
+vpp-clab# trace add af-packet-input 100
+vpp-clab# wait 3
+vpp-clab# show trace
+------------------- Start of thread 0 vpp_main -------------------
+Packet 1
+
+00:07:03:979275: af-packet-input
+  af_packet: hw_if_index 1 rx-queue 0 next-index 4
+    block 47:
+      address 0x7fbf23b7d000 version 2 seq_num 48 pkt_num 0
+    tpacket3_hdr:
+      status 0x20000001 len 98 snaplen 98 mac 92 net 106
+      sec 0x68164381 nsec 0x258e7659 vlan 0 vlan_tpid 0
+    vnet-hdr:
+      flags 0x00 gso_type 0x00 hdr_len 0
+      gso_size 0 csum_start 0 csum_offset 0
+00:07:03:979293: ethernet-input
+  IP4: 02:42:09:97:28:c6 -> 02:42:c0:00:02:02
+00:07:03:979306: ip4-input
+  ICMP: 192.0.2.1 -> 192.0.2.2
+    tos 0x00, ttl 64, length 84, checksum 0x5e92 dscp CS0 ecn NON_ECN
+    fragment id 0x5813, flags DONT_FRAGMENT
+  ICMP echo_request checksum 0xc16 id 21197
+00:07:03:979315: ip4-lookup
+  fib 0 dpo-idx 9 flow hash: 0x00000000
+  ICMP: 192.0.2.1 -> 192.0.2.2
+    tos 0x00, ttl 64, length 84, checksum 0x5e92 dscp CS0 ecn NON_ECN
+    fragment id 0x5813, flags DONT_FRAGMENT
+  ICMP echo_request checksum 0xc16 id 21197
+00:07:03:979322: ip4-receive
+    fib:0 adj:9 flow:0x00000000
+  ICMP: 192.0.2.1 -> 192.0.2.2
+    tos 0x00, ttl 64, length 84, checksum 0x5e92 dscp CS0 ecn NON_ECN
+    fragment id 0x5813, flags DONT_FRAGMENT
+  ICMP echo_request checksum 0xc16 id 21197
+00:07:03:979323: ip4-icmp-input
+  ICMP: 192.0.2.1 -> 192.0.2.2
+    tos 0x00, ttl 64, length 84, checksum 0x5e92 dscp CS0 ecn NON_ECN
+    fragment id 0x5813, flags DONT_FRAGMENT
+  ICMP echo_request checksum 0xc16 id 21197
+00:07:03:979323: ip4-icmp-echo-request
+  ICMP: 192.0.2.1 -> 192.0.2.2
+    tos 0x00, ttl 64, length 84, checksum 0x5e92 dscp CS0 ecn NON_ECN
+    fragment id 0x5813, flags DONT_FRAGMENT
+  ICMP echo_request checksum 0xc16 id 21197
+00:07:03:979326: ip4-load-balance
+  fib 0 dpo-idx 5 flow hash: 0x00000000
+  ICMP: 192.0.2.2 -> 192.0.2.1
+    tos 0x00, ttl 64, length 84, checksum 0x88e1 dscp CS0 ecn NON_ECN
+    fragment id 0x2dc4, flags DONT_FRAGMENT
+  ICMP echo_reply checksum 0x1416 id 21197
+00:07:03:979325: ip4-rewrite
+  tx_sw_if_index 1 dpo-idx 5 : ipv4 via 192.0.2.1 eth1: mtu:1500 next:3 flags:[] 0242099728c60242c00002020800 flow hash: 0x00000000
+  00000000: 0242099728c60242c00002020800450000542dc44000400188e1c0000202c000
+  00000020: 02010000141652cd00018143166800000000399d0900000000001011
+00:07:03:979326: eth1-output
+  eth1 flags 0x02180005
+  IP4: 02:42:c0:00:02:02 -> 02:42:09:97:28:c6
+  ICMP: 192.0.2.2 -> 192.0.2.1
+    tos 0x00, ttl 64, length 84, checksum 0x88e1 dscp CS0 ecn NON_ECN
+    fragment id 0x2dc4, flags DONT_FRAGMENT
+  ICMP echo_reply checksum 0x1416 id 21197
+00:07:03:979327: eth1-tx
+  af_packet: hw_if_index 1 tx-queue 0
+    tpacket3_hdr:
+      status 0x1 len 108 snaplen 108 mac 0 net 0
+      sec 0x0 nsec 0x0 vlan 0 vlan_tpid 0
+    vnet-hdr:
+      flags 0x00 gso_type 0x00 hdr_len 0
+      gso_size 0 csum_start 0 csum_offset 0
+    buffer 0xf97c4:
+      current data 0, length 98, buffer-pool 0, ref-count 1, trace handle 0x0
+      local l2-hdr-offset 0 l3-hdr-offset 14 
+    IP4: 02:42:c0:00:02:02 -> 02:42:09:97:28:c6
+    ICMP: 192.0.2.2 -> 192.0.2.1
+      tos 0x00, ttl 64, length 84, checksum 0x88e1 dscp CS0 ecn NON_ECN
+      fragment id 0x2dc4, flags DONT_FRAGMENT
+    ICMP echo_reply checksum 0x1416 id 21197
+```
+
+Well, that's a mouthfull, isn't it! Here, I get to show you VPP in action. After receiving the
+packet on its `af-packet-input` node from 192.0.2.1 (Summer, who is pinging us) to 192.0.2.2 (the
+VPP container), the packet traverses the dataplane graph. It goes through `ethernet-input`, then
+`ip4-input`, which sees it's destined to an IPv4 address configured, so the packet is handed to
+`ip4-receive`. That one sees that the IP protocol is ICMP, so it hands the packet to
+`ip4-icmp-input` which notices that the packet is an ICMP echo request, so off to
+`ip4-icmp-echo-request` our little packet goes. The ICMP plugin in VPP now answers by
+`ip4-rewrite`'ing the packet, sending the return to 192.0.2.1 at MAC address `02:42:09:97:28:c6`
+(this is Summer, the host doing the pinging!), after which the newly created ICMP echo-reply is
+handed to `eth1-output` which marshalls it back into the kernel's AF_PACKET interface using
+`eth1-tx`.
+
+Boom. I could not be more pleased.
+
+## What's Next
+
+This was a nice exercise for me! I'm going this direction becaue the
+[[Containerlab](https://containerlab.dev)] framework will start containers with given NOS images, 
+not too dissimilar from the one I just made, and then attaches `veth` pairs between the containers.
+I started dabbling with a [[pull-request](https://github.com/srl-labs/containerlab/pull/2571)], but
+I got stuck with a part of the Containerlab code that pre-deploys config files into the containers.
+You see, I will need to generate two files:
+
+1.   A `startup.conf` file that is specific to the containerlab Docker container. I'd like them to
+     each set their own hostname so that the CLI has a unique prompt. I can do this by setting `unix
+     { cli-prompt {{ .ShortName }}# }` in the template renderer.
+1.   Containerlab will know all of the veth pairs that are planned to be created into each VPP
+     container. I'll need it to then write a little snippet of config that does the `create
+     host-interface` spiel, to attach these `veth` pairs to the VPP dataplane.
+
+I reached out to Roman from Nokia, who is one of the authors and current maintainer of Containerlab.
+Roman was keen to help out, and seeing as he knows the COntainerlab stuff well, and I know the VPP
+stuff well, this is a reasonable partnership! Soon, he and I plan to have a bare-bones setup that
+will connect a few VPP containers together with an SR Linux node in a lab. Stand by!
+
+Once we have that, there's still quite some work for me to do. Notably:
+*    Configuration persistence. `clab` allows you to save the running config. For that, I'll need to
+     introduce [[vppcfg](https://git.ipng.ch/ipng/vppcfg)] and a means to invoke it when
+     the lab operator wants to save their config, and then reconfigure VPP when the container
+     restarts.
+*    I'll need to have a few files from `clab` shared with the host, notably the `startup.conf` and
+     `vppcfg.yaml`, as well as some manual pre- and post-flight configuration for the more esoteric
+     stuff. Building the plumbing for this is a TODO for now.
+
+## Acknowledgements
+
+I wanted to give a shout-out to Nardus le Roux who inspired me to contribute this Containerlab VPP
+node type, and to Roman Dodin for his help getting the Containerlab parts squared away when I got a
+little bit stuck.
+
+First order of business: get it to ping at all ... it'll go faster from there on out :)

content/articles/2025-05-04-containerlab-2.md

@@ -0,0 +1,373 @@
+---
+date: "2025-05-04T15:07:23Z"
+title: 'VPP in Containerlab - Part 2'
+params:
+  asciinema: true
+---
+
+{{< image float="right" src="/assets/containerlab/containerlab.svg" alt="Containerlab Logo" width="12em" >}}
+
+# Introduction
+
+From time to time the subject of containerized VPP instances comes up. At IPng, I run the routers in
+AS8298 on bare metal (Supermicro and Dell hardware), as it allows me to maximize performance.
+However, VPP is quite friendly in virtualization. Notably, it runs really well on virtual machines
+like Qemu/KVM or VMWare. I can pass through PCI devices directly to the host, and use CPU pinning to
+allow the guest virtual machine access to the underlying physical hardware. In such a mode, VPP
+performance almost the same as on bare metal. But did you know that VPP can also run in Docker?
+
+The other day I joined the [[ZANOG'25](https://nog.net.za/event1/zanog25/)] in Durban, South Africa.
+One of the presenters was Nardus le Roux of Nokia, and he showed off a project called
+[[Containerlab](https://containerlab.dev/)], which provides a CLI for orchestrating and managing
+container-based networking labs. It starts the containers, builds virtual wiring between them to
+create lab topologies of users' choice and manages the lab lifecycle.
+
+Quite regularly I am asked 'when will you add VPP to Containerlab?', but at ZANOG I made a promise
+to actually add it. In my previous [[article]({{< ref 2025-05-03-containerlab-1.md >}})], I took
+a good look at VPP as a dockerized container. In this article, I'll explore how to make such a
+container run in Containerlab!
+
+## Completing the Docker container
+
+Just having VPP running by itself in a container is not super useful (although it _is_ cool!). I
+decide first to add a few bits and bobs that will come in handy in the `Dockerfile`:
+
+```
+FROM debian:bookworm
+ARG DEBIAN_FRONTEND=noninteractive
+ARG VPP_INSTALL_SKIP_SYSCTL=true
+ARG REPO=release
+EXPOSE 22/tcp
+RUN apt-get update && apt-get -y install curl procps tcpdump iproute2 iptables \
+  iputils-ping net-tools git python3 python3-pip vim-tiny openssh-server bird2 \
+  mtr-tiny traceroute && apt-get clean
+
+# Install VPP
+RUN mkdir -p /var/log/vpp /root/.ssh/
+RUN curl -s https://packagecloud.io/install/repositories/fdio/${REPO}/script.deb.sh |  bash
+RUN apt-get update && apt-get -y install vpp vpp-plugin-core && apt-get clean
+
+# Build vppcfg
+RUN pip install --break-system-packages build netaddr yamale argparse pyyaml ipaddress
+RUN git clone https://git.ipng.ch/ipng/vppcfg.git && cd vppcfg && python3 -m build && \
+    pip install --break-system-packages dist/vppcfg-*-py3-none-any.whl
+
+# Config files
+COPY files/etc/vpp/* /etc/vpp/
+COPY files/etc/bird/* /etc/bird/
+COPY files/init-container.sh /sbin/
+RUN chmod 755 /sbin/init-container.sh
+CMD ["/sbin/init-container.sh"]
+```
+
+A few notable additions:
+*   ***vppcfg*** is a handy utility I wrote and discussed in a previous [[article]({{< ref
+    2022-04-02-vppcfg-2 >}})]. Its purpose is to take YAML file that describes the configuration of
+    the dataplane (like which interfaces, sub-interfaces, MTU, IP addresses and so on), and then
+    apply this safely to a running dataplane. You can check it out in my
+    [[vppcfg](https://git.ipng.ch/ipng/vppcfg)] git repository.
+*   ***openssh-server*** will come in handy to log in to the container, in addition to the already
+    available `docker exec`.
+*   ***bird2*** which will be my controlplane of choice. At a future date, I might also add FRR,
+    which may be a good alterantive for some. VPP works well with both. You can check out Bird on
+    the nic.cz [[website](https://bird.network.cz/?get_doc&f=bird.html&v=20)].
+
+I'll add a couple of default config files for Bird and VPP, and replace the CMD with a generic
+`/sbin/init-container.sh` in which I can do any late binding stuff before launching VPP.
+
+### Initializing the Container
+
+#### VPP Containerlab: NetNS
+
+VPP's Linux Control Plane plugin wants to run in its own network namespace. So the first order of
+business of `/sbin/init-container.sh` is to create it:
+
+```
+NETNS=${NETNS:="dataplane"}
+
+echo "Creating dataplane namespace"
+/usr/bin/mkdir -p /etc/netns/$NETNS
+/usr/bin/touch /etc/netns/$NETNS/resolv.conf
+/usr/sbin/ip netns add $NETNS
+```
+
+#### VPP Containerlab: SSH
+
+Then, I'll set the root password (which is `vpp` by the way), and start aan SSH daemon which allows
+for password-less logins:
+
+```
+echo "Starting SSH, with credentials root:vpp"
+sed -i -e 's,^#PermitRootLogin prohibit-password,PermitRootLogin yes,' /etc/ssh/sshd_config
+sed -i -e 's,^root:.*,root:$y$j9T$kG8pyZEVmwLXEtXekQCRK.$9iJxq/bEx5buni1hrC8VmvkDHRy7ZMsw9wYvwrzexID:20211::::::,' /etc/shadow
+/etc/init.d/ssh start
+```
+
+#### VPP Containerlab: Bird2
+
+I can already predict that Bird2 won't be the only option for a controlplane, even though I'm a huge
+fan of it. Therefore, I'll make it configurable to leave the door open for other controlplane
+implementations in the future:
+
+```
+BIRD_ENABLED=${BIRD_ENABLED:="true"}
+
+if [ "$BIRD_ENABLED" == "true" ]; then
+  echo "Starting Bird in $NETNS"
+  mkdir -p /run/bird /var/log/bird
+  chown bird:bird /var/log/bird
+  ROUTERID=$(ip -br a show eth0 | awk '{ print $3 }' | cut -f1 -d/)
+  sed -i -e "s,.*router id .*,router id $ROUTERID; # Set by container-init.sh," /etc/bird/bird.conf
+  /usr/bin/nsenter --net=/var/run/netns/$NETNS /usr/sbin/bird -u bird -g bird
+fi
+```
+
+I am reminded that Bird won't start if it cannot determine its _router id_. When I start it in the
+`dataplane` namespace, it will immediately exit, because there will be no IP addresses configured
+yet. But luckily, it logs its complaint and it's easily addressed. I decide to take the management
+IPv4 address from `eth0` and write that into the `bird.conf` file, which otherwise does some basic
+initialization that I described in a previous [[article]({{< ref 2021-09-02-vpp-5 >}})], so I'll
+skip that here. However, I do include an empty file called `/etc/bird/bird-local.conf` for users to
+further configure Bird2.
+
+#### VPP Containerlab: Binding veth pairs
+
+When Containerlab starts the VPP container, it'll offer it a set of `veth` ports that connect this
+container to other nodes in the lab. This is done by the `links` list in the topology file
+[[ref](https://containerlab.dev/manual/network/)]. It's my goal to take all of the interfaces
+that are of type `veth`, and generate a little snippet to grab them and bind them into VPP while
+setting their MTU to 9216 to allow for jumbo frames:
+
+```
+CLAB_VPP_FILE=${CLAB_VPP_FILE:=/etc/vpp/clab.vpp}
+
+echo "Generating $CLAB_VPP_FILE"
+: > $CLAB_VPP_FILE
+MTU=9216
+for IFNAME in $(ip -br link show type veth | cut -f1 -d@ | grep -v '^eth0$' | sort); do
+  MAC=$(ip -br link show dev $IFNAME | awk '{ print $3 }')
+  echo " * $IFNAME hw-addr $MAC mtu $MTU"
+  ip link set $IFNAME up mtu $MTU
+  cat << EOF >> $CLAB_VPP_FILE
+create host-interface name $IFNAME hw-addr $MAC
+set interface name host-$IFNAME $IFNAME
+set interface mtu $MTU $IFNAME
+set interface state $IFNAME up
+
+EOF
+done
+```
+
+{{< image width="5em" float="left" src="/assets/shared/warning.png" alt="Warning" >}}
+
+One thing I realized is that VPP will assign a random MAC address on its copy of the `veth` port,
+which is not great. I'll explicitly configure it with the same MAC address as the `veth` interface
+itself, otherwise I'd have to put the interface into promiscuous mode.
+
+#### VPP Containerlab: VPPcfg
+
+I'm almost ready, but I have one more detail. The user will be able to offer a
+[[vppcfg](https://git.ipng.ch/ipng/vppcfg)] YAML file to configure the interfaces and so on. If such
+a file exists, I'll apply it to the dataplane upon startup:
+
+```
+VPPCFG_VPP_FILE=${VPPCFG_VPP_FILE:=/etc/vpp/vppcfg.vpp}
+
+echo "Generating $VPPCFG_VPP_FILE"
+: > $VPPCFG_VPP_FILE
+if [ -r /etc/vpp/vppcfg.yaml ]; then
+  vppcfg plan --novpp -c /etc/vpp/vppcfg.yaml -o $VPPCFG_VPP_FILE
+fi
+```
+
+Once the VPP process starts, it'll execute `/etc/vpp/bootstrap.vpp`, which in turn executes these
+newly generated `/etc/vpp/clab.vpp` to grab the `veth` interfaces, and then `/etc/vpp/vppcfg.vpp` to
+further configure the dataplane. Easy peasy!
+
+### Adding VPP to Containerlab
+
+Roman points out a previous integration for the 6WIND VSR in
+[[PR#2540](https://github.com/srl-labs/containerlab/pull/2540)]. This serves as a useful guide to
+get me started. I fork the repo, create a branch so that Roman can also add a few commits, and
+together we start hacking in [[PR#2571](https://github.com/srl-labs/containerlab/pull/2571)].
+
+First, I add the documentation skeleton in `docs/manual/kinds/fdio_vpp.md`, which links in from a
+few other places, and will be where the end-user facing documentation will live. That's about half
+the contributed LOC, right there!
+
+Next, I'll create a Go module in `nodes/fdio_vpp/fdio_vpp.go` which doesn't do much other than
+creating the `struct`, and its required `Register` and `Init` functions. The `Init` function ensures
+the right capabilities are set in Docker, and the right devices are bound for the container. 
+
+I notice that Containerlab rewrites the Dockerfile `CMD` string and prepends an `if-wait.sh` script
+to it. This is because when Containerlab starts the container, it'll still be busy adding these
+`link` interfaces to it, and if a container starts too quickly, it may not see all the interfaces.
+So, containerlab informs the container using an environment variable called `CLAB_INTFS`, so this
+script simply sleeps for a while until that exact amount of interfaces are present. Ok, cool beans.
+
+Roman helps me a bit with Go templating. You see, I think it'll be slick to have the CLI prompt for
+the VPP containers to reflect their hostname, because normally, VPP will assign `vpp# `. I add the
+template in `nodes/fdio_vpp/vpp_startup_config.go.tpl` and it only has one variable expansion: `unix
+{ cli-prompt {{ .ShortName }}# }`. But I totally think it's worth it, because when running many VPP
+containers in the lab, it could otherwise get confusing. 
+
+Roman also shows me a trick in the function `PostDeploy()`, which will write the user's SSH pubkeys
+to `/root/.ssh/authorized_keys`. This allows users to log in without having to use password
+authentication. 
+
+Collectively, we decide to punt on the `SaveConfig` function until we're a bit further along. I have
+an idea how this would work, basically along the lines of calling `vppcfg dump` and bind-mounting
+that file into the lab directory somewhere. This way, upon restarting, the YAML file can be re-read
+and the dataplane initialized. But it'll be for another day.
+
+After the main module is finished, all I have to do is add it to `clab/register.go` and that's just
+about it. In about 170 lines of code, 50 lines of Go template, and 170 lines of Markdown, this
+contribution is about ready to ship!
+
+### Containerlab: Demo
+
+After I finish writing the documentation, I decide to include a demo with a quickstart to help folks
+along. A simple lab showing two VPP instances and two Alpine Linux clients can be found on
+[[git.ipng.ch/ipng/vpp-containerlab](https://git.ipng.ch/ipng/vpp-containerlab)]. Simply check out the
+repo and start the lab, like so:
+
+```
+$ git clone https://git.ipng.ch/ipng/vpp-containerlab.git
+$ cd vpp-containerlab
+$ containerlab deploy --topo vpp.clab.yml
+```
+
+#### Containerlab: configs
+
+The file `vpp.clab.yml` contains an example topology existing of two VPP instances connected each to
+one Alpine linux container, in the following topology:
+
+{{< image src="/assets/containerlab/learn-vpp.png" alt="Containerlab Topo" width="100%" >}}
+
+Two relevant files for each VPP router are included in this
+[[repository](https://git.ipng.ch/ipng/vpp-containerlab)]:
+1.   `config/vpp*/vppcfg.yaml` configures the dataplane interfaces, including a loopback address.
+1.   `config/vpp*/bird-local.conf` configures the controlplane to enable BFD and OSPF.
+
+To illustrate these files, let me take a closer look at node `vpp1`. It's VPP dataplane
+configuration looks like this:
+```
+pim@summer:~/src/vpp-containerlab$ cat config/vpp1/vppcfg.yaml 
+interfaces:
+  eth1:
+    description: 'To client1'
+    mtu: 1500
+    lcp: eth1
+    addresses: [ 10.82.98.65/28, 2001:db8:8298:101::1/64 ]
+  eth2:
+    description: 'To vpp2'
+    mtu: 9216
+    lcp: eth2
+    addresses: [ 10.82.98.16/31, 2001:db8:8298:1::1/64 ]
+loopbacks:
+  loop0:
+    description: 'vpp1'
+    lcp: loop0
+    addresses: [ 10.82.98.0/32, 2001:db8:8298::/128 ]
+```
+
+Then, I enable BFD, OSPF and OSPFv3 on `eth2` and `loop0` on both of the VPP routers:
+```
+pim@summer:~/src/vpp-containerlab$ cat config/vpp1/bird-local.conf 
+protocol bfd bfd1 {
+  interface "eth2" { interval 100 ms; multiplier 30; };
+}
+
+protocol ospf v2 ospf4 {
+  ipv4 { import all; export all; };
+  area 0 {
+    interface "loop0" { stub yes; };
+    interface "eth2" { type pointopoint; cost 10; bfd on; };
+  };
+}
+
+protocol ospf v3 ospf6 {
+  ipv6 { import all; export all; };
+  area 0 {
+    interface "loop0" { stub yes; };
+    interface "eth2" { type pointopoint; cost 10; bfd on; };
+  };
+}
+```
+
+#### Containerlab: playtime!
+
+Once the lab comes up, I can SSH to the VPP containers (`vpp1` and `vpp2`) which have my SSH pubkeys
+installed thanks to Roman's work. Barring that, I could still log in as user `root` using
+password `vpp`. VPP runs its own network namespace called `dataplane`, which is very similar to SR
+Linux default `network-instance`. I can join that namespace to take a closer look:
+
+```
+pim@summer:~/src/vpp-containerlab$ ssh root@vpp1
+root@vpp1:~# nsenter --net=/var/run/netns/dataplane
+root@vpp1:~# ip -br a
+lo               DOWN           
+loop0            UP             10.82.98.0/32 2001:db8:8298::/128 fe80::dcad:ff:fe00:0/64 
+eth1             UNKNOWN        10.82.98.65/28 2001:db8:8298:101::1/64 fe80::a8c1:abff:fe77:acb9/64 
+eth2             UNKNOWN        10.82.98.16/31 2001:db8:8298:1::1/64 fe80::a8c1:abff:fef0:7125/64 
+
+root@vpp1:~# ping 10.82.98.1
+PING 10.82.98.1 (10.82.98.1) 56(84) bytes of data.
+64 bytes from 10.82.98.1: icmp_seq=1 ttl=64 time=9.53 ms
+64 bytes from 10.82.98.1: icmp_seq=2 ttl=64 time=15.9 ms
+^C
+--- 10.82.98.1 ping statistics ---
+2 packets transmitted, 2 received, 0% packet loss, time 1002ms
+rtt min/avg/max/mdev = 9.530/12.735/15.941/3.205 ms
+```
+
+From `vpp1`, I can tell that Bird2's OSPF adjacency has formed, because I can ping the `loop0`
+address of `vpp2` router on 10.82.98.1. Nice! The two client nodes are running a minimalistic Alpine
+Linux container, which doesn't ship with SSH by default. But of course I can still enter the
+containers using `docker exec`, like so:
+
+```
+pim@summer:~/src/vpp-containerlab$ docker exec -it client1 sh
+/ # ip addr show dev eth1
+531235: eth1@if531234: <BROADCAST,MULTICAST,UP,LOWER_UP,M-DOWN> mtu 9500 qdisc noqueue state UP 
+    link/ether 00:c1:ab:00:00:01 brd ff:ff:ff:ff:ff:ff
+    inet 10.82.98.66/28 scope global eth1
+       valid_lft forever preferred_lft forever
+    inet6 2001:db8:8298:101::2/64 scope global 
+       valid_lft forever preferred_lft forever
+    inet6 fe80::2c1:abff:fe00:1/64 scope link 
+       valid_lft forever preferred_lft forever
+/ # traceroute 10.82.98.82
+traceroute to 10.82.98.82 (10.82.98.82), 30 hops max, 46 byte packets
+ 1  10.82.98.65 (10.82.98.65)  5.906 ms  7.086 ms  7.868 ms
+ 2  10.82.98.17 (10.82.98.17)  24.007 ms  23.349 ms  15.933 ms
+ 3  10.82.98.82 (10.82.98.82)  39.978 ms  31.127 ms  31.854 ms
+
+/ # traceroute 2001:db8:8298:102::2
+traceroute to 2001:db8:8298:102::2 (2001:db8:8298:102::2), 30 hops max, 72 byte packets
+ 1  2001:db8:8298:101::1 (2001:db8:8298:101::1)  0.701 ms  7.144 ms  7.900 ms
+ 2  2001:db8:8298:1::2 (2001:db8:8298:1::2)  23.909 ms  22.943 ms  23.893 ms
+ 3  2001:db8:8298:102::2 (2001:db8:8298:102::2)  31.964 ms  30.814 ms  32.000 ms
+```
+
+From the vantage point of `client1`, the first hop represents the `vpp1` node, which forwards to
+`vpp2`, which finally forwards to `client2`, which shows that both VPP routers are passing traffic.
+Dope!
+
+## Results
+
+After all of this deep-diving, all that's left is for me to demonstrate the Containerlab by means of
+this little screencast [[asciinema](/assets/containerlab/vpp-containerlab.cast)].  I hope you enjoy
+it as much as I enjoyed creating it:
+
+{{< asciinema src="/assets/containerlab/vpp-containerlab.cast" >}}
+
+## Acknowledgements
+
+I wanted to give a shout-out Roman Dodin for his help getting the Containerlab parts squared away
+when I got a little bit stuck. He took the time to explain the internals and idiom of Containerlab
+project, which really saved me a tonne of time. He also pair-programmed the
+[[PR#2471](https://github.com/srl-labs/containerlab/pull/2571)] with me over the span of two
+evenings.
+
+Collaborative open source rocks!

hugo.yaml

@@ -0,0 +1,38 @@
+baseURL: 'https://ipng.ch/'
+languageCode: 'en-us'
+title: "IPng Networks"
+theme: 'hugo-theme-ipng'
+
+mainSections: ["articles"]
+
+params:
+  author: "IPng Networks GmbH"
+  siteHeading: "IPng Networks"
+  favicon: "favicon.ico"
+  showBlogLatest: false
+  mainSections: ["articles"]
+  showTaxonomyLinks: false
+  nBlogLatest: 14 # number of blog post om the home page
+  Paginate: 30
+  blogLatestHeading: "Latest Dabblings"
+  footer: "Copyright 2021- IPng Networks GmbH, all rights reserved"
+
+  social:
+    email: "info+www@ipng.ch"
+    mastodon: "@IPngNetworks"
+    twitter: "IPngNetworks"
+    linkedin: "pimvanpelt"
+    github: "pimvanpelt"
+    instagram: "IPngNetworks"
+    rss: true
+
+taxonomies:
+  year: "year"
+  month: "month"
+  tags: "tags"
+  categories: "categories"
+
+permalinks:
+  articles: "/s/articles/:year/:month/:day/:slug"
+
+ignoreLogs: [ "warning-goldmark-raw-html" ]

static/.well-known/security.txt

@@ -0,0 +1,5 @@
+Canonical: https://ipng.ch/.well-known/security.txt
+Expires: 2026-01-01T00:00:00.000Z
+Contact: mailto:info@ipng.ch
+Contact: https://ipng.ch/s/contact/
+Preferred-Languages: en, nl, de

static/app/go/index.html

@@ -0,0 +1,55 @@
+<!DOCTYPE html>
+<html lang="en-us">
+ <head>
+  <title>Javascript Redirector for RFID / NFC / nTAG</title>
+  <meta name="robots" content="noindex,nofollow">
+  <meta charset="utf-8">
+  <script type="text/JavaScript">
+
+const ntag_list = [
+  "/s/articles/2021/09/21/vpp-linux-cp-part7/",
+  "/s/articles/2021/12/23/vpp-linux-cp-virtual-machine-playground/",
+  "/s/articles/2022/01/12/case-study-virtual-leased-line-vll-in-vpp/",
+  "/s/articles/2022/02/14/case-study-vlan-gymnastics-with-vpp/",
+  "/s/articles/2022/03/27/vpp-configuration-part1/",
+  "/s/articles/2022/10/14/vpp-lab-setup/",
+  "/s/articles/2023/03/11/case-study-centec-mpls-core/",
+  "/s/articles/2023/04/09/vpp-monitoring/",
+  "/s/articles/2023/05/28/vpp-mpls-part-4/",
+  "/s/articles/2023/11/11/debian-on-mellanox-sn2700-32x100g/",
+  "/s/articles/2023/12/17/debian-on-ipngs-vpp-routers/",
+  "/s/articles/2024/01/27/vpp-python-api/",
+  "/s/articles/2024/02/10/vpp-on-freebsd-part-1/",
+  "/s/articles/2024/03/06/vpp-with-babel-part-1/",
+  "/s/articles/2024/04/06/vpp-with-loopback-only-ospfv3-part-1/",
+  "/s/articles/2024/04/27/freeix-remote/"
+];
+
+var redir_url = "https://ipng.ch/";
+var key = window.location.hash.slice(1);
+if (key.startsWith("ntag")) {
+  let week = Math.round(new Date().getTime() / 1000 / (7*24*3400));
+  let num = parseInt(key.slice(-2));
+  let idx = (num + week) % ntag_list.length;
+  console.log("(ntag " + num + " + week number " + week + ") % " + ntag_list.length + " = " + idx);
+  redir_url = ntag_list[idx];
+}
+
+console.log("Redirecting to " + redir_url + " - off you go!");
+window.location = redir_url;
+  </script>
+ </head>
+ <body>
+
+<pre>
+Usage: https://ipng.ch/app/go/#&lt;key&gt;
+Example: <a href="/app/go/#ntag00">#ntag00</a>
+
+Also, this page requires javascript.
+
+Love,
+  IPng Networks.
+</pre>
+
+ </body>
+</html>

static/assets/frys-ix/arista-leaf.conf

@@ -0,0 +1,169 @@
+no aaa root
+!
+hardware counter feature vtep decap
+hardware counter feature vtep encap
+!
+service routing protocols model multi-agent
+!
+hostname arista-leaf
+!
+router l2-vpn
+   arp learning bridged
+!
+spanning-tree mode mstp
+!
+system l1
+   unsupported speed action error
+   unsupported error-correction action error
+!
+vlan 2604
+   name v-peeringlan
+!
+interface Ethernet1/1
+!
+interface Ethernet2/1
+!
+interface Ethernet3/1
+!
+interface Ethernet4/1
+!
+interface Ethernet5/1
+!
+interface Ethernet6/1
+!
+interface Ethernet7/1
+!
+interface Ethernet8/1
+!
+interface Ethernet9/1
+   shutdown
+   speed forced 10000full
+!
+interface Ethernet9/2
+   shutdown
+!
+interface Ethernet9/3
+   speed forced 10000full
+   switchport access vlan 2604
+!
+interface Ethernet9/4
+   shutdown
+!
+interface Ethernet10/1
+!
+interface Ethernet10/2
+   shutdown
+!
+interface Ethernet10/4
+   shutdown
+!
+interface Ethernet11/1
+!
+interface Ethernet12/1
+!
+interface Ethernet13/1
+!
+interface Ethernet14/1
+!
+interface Ethernet15/1
+!
+interface Ethernet16/1
+!
+interface Ethernet17/1
+!
+interface Ethernet18/1
+!
+interface Ethernet19/1
+!
+interface Ethernet20/1
+!
+interface Ethernet21/1
+!
+interface Ethernet22/1
+!
+interface Ethernet23/1
+!
+interface Ethernet24/1
+!
+interface Ethernet25/1
+!
+interface Ethernet26/1
+!
+interface Ethernet27/1
+!
+interface Ethernet28/1
+!
+interface Ethernet29/1
+   no switchport
+!
+interface Ethernet30/1
+   load-interval 1
+   mtu 9190
+   no switchport
+   ip address 198.19.17.10/31
+   ip ospf cost 10
+   ip ospf network point-to-point
+   ip ospf area 0.0.0.0
+!
+interface Ethernet31/1
+   load-interval 1
+   mtu 9190
+   no switchport
+   ip address 198.19.17.3/31
+   ip ospf cost 1000
+   ip ospf network point-to-point
+   ip ospf area 0.0.0.0
+!
+interface Ethernet32/1
+   load-interval 1
+   mtu 9190
+   no switchport
+   ip address 198.19.17.5/31
+   ip ospf cost 1000
+   ip ospf network point-to-point
+   ip ospf area 0.0.0.0
+!
+interface Loopback0
+   ip address 198.19.16.2/32
+   ip ospf area 0.0.0.0
+!
+interface Loopback1
+   ip address 198.19.18.2/32
+!
+interface Management1
+   ip address dhcp
+!
+interface Vxlan1
+   vxlan source-interface Loopback1
+   vxlan udp-port 4789
+   vxlan vlan 2604 vni 2604
+!
+ip routing
+!
+ip route 0.0.0.0/0 Management1 10.75.8.1
+!
+router bgp 65500
+   neighbor evpn peer group
+   neighbor evpn remote-as 65500
+   neighbor evpn update-source Loopback0
+   neighbor evpn ebgp-multihop 3
+   neighbor evpn send-community extended
+   neighbor evpn maximum-routes 12000 warning-only
+   neighbor 198.19.16.0 peer group evpn
+   neighbor 198.19.16.1 peer group evpn
+   !
+   vlan 2604
+      rd 65500:2604
+      route-target both 65500:2604
+      redistribute learned
+   !
+   address-family evpn
+      neighbor evpn activate
+!
+router ospf 65500
+   router-id 198.19.16.2
+   redistribute connected
+   network 198.19.0.0/16 area 0.0.0.0
+   max-lsa 12000
+!
+end

static/assets/frys-ix/equinix.conf

@@ -0,0 +1,90 @@
+set / interface ethernet-1/1 admin-state disable
+set / interface ethernet-1/9 admin-state enable
+set / interface ethernet-1/9 breakout-mode num-breakout-ports 4
+set / interface ethernet-1/9 breakout-mode breakout-port-speed 10G
+set / interface ethernet-1/9/3 admin-state enable
+set / interface ethernet-1/9/3 vlan-tagging true
+set / interface ethernet-1/9/3 subinterface 0 type bridged
+set / interface ethernet-1/9/3 subinterface 0 admin-state enable
+set / interface ethernet-1/9/3 subinterface 0 vlan encap untagged
+set / interface ethernet-1/29 admin-state enable
+set / interface ethernet-1/29 subinterface 0 type routed
+set / interface ethernet-1/29 subinterface 0 admin-state enable
+set / interface ethernet-1/29 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/29 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/29 subinterface 0 ipv4 address 198.19.17.0/31
+set / interface ethernet-1/29 subinterface 0 ipv6 admin-state enable
+set / interface lo0 admin-state enable
+set / interface lo0 subinterface 0 admin-state enable
+set / interface lo0 subinterface 0 ipv4 admin-state enable
+set / interface lo0 subinterface 0 ipv4 address 198.19.16.0/32
+set / interface mgmt0 admin-state enable
+set / interface mgmt0 subinterface 0 admin-state enable
+set / interface mgmt0 subinterface 0 ipv4 admin-state enable
+set / interface mgmt0 subinterface 0 ipv4 dhcp-client
+set / interface mgmt0 subinterface 0 ipv6 admin-state enable
+set / interface mgmt0 subinterface 0 ipv6 dhcp-client
+set / interface system0 admin-state enable
+set / interface system0 subinterface 0 admin-state enable
+set / interface system0 subinterface 0 ipv4 admin-state enable
+set / interface system0 subinterface 0 ipv4 address 198.19.18.0/32
+set / network-instance default type default
+set / network-instance default admin-state enable
+set / network-instance default description "fabric: dc2 role: spine"
+set / network-instance default router-id 198.19.16.0
+set / network-instance default ip-forwarding receive-ipv4-check false
+set / network-instance default interface ethernet-1/29.0
+set / network-instance default interface lo0.0
+set / network-instance default interface system0.0
+set / network-instance default protocols bgp admin-state enable
+set / network-instance default protocols bgp autonomous-system 65500
+set / network-instance default protocols bgp router-id 198.19.16.0
+set / network-instance default protocols bgp dynamic-neighbors accept match 198.19.16.0/24 peer-group overlay
+set / network-instance default protocols bgp afi-safi evpn admin-state enable
+set / network-instance default protocols bgp preference ibgp 170
+set / network-instance default protocols bgp route-advertisement rapid-withdrawal true
+set / network-instance default protocols bgp route-advertisement wait-for-fib-install false
+set / network-instance default protocols bgp group overlay peer-as 65500
+set / network-instance default protocols bgp group overlay afi-safi evpn admin-state enable
+set / network-instance default protocols bgp group overlay afi-safi ipv4-unicast admin-state disable
+set / network-instance default protocols bgp group overlay local-as as-number 65500
+set / network-instance default protocols bgp group overlay route-reflector client true
+set / network-instance default protocols bgp group overlay transport local-address 198.19.16.0
+set / network-instance default protocols bgp neighbor 198.19.16.1 admin-state enable
+set / network-instance default protocols bgp neighbor 198.19.16.1 peer-group overlay
+set / network-instance default protocols ospf instance default admin-state enable
+set / network-instance default protocols ospf instance default version ospf-v2
+set / network-instance default protocols ospf instance default router-id 198.19.16.0
+set / network-instance default protocols ospf instance default export-policy ospf
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/29.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface lo0.0
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface system0.0
+set / network-instance mgmt type ip-vrf
+set / network-instance mgmt admin-state enable
+set / network-instance mgmt description "Management network instance"
+set / network-instance mgmt interface mgmt0.0
+set / network-instance mgmt protocols linux import-routes true
+set / network-instance mgmt protocols linux export-routes true
+set / network-instance mgmt protocols linux export-neighbors true
+set / network-instance peeringlan type mac-vrf
+set / network-instance peeringlan admin-state enable
+set / network-instance peeringlan interface ethernet-1/9/3.0
+set / network-instance peeringlan vxlan-interface vxlan1.2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 admin-state enable
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 vxlan-interface vxlan1.2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 evi 2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 routes bridge-table mac-ip advertise true
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-distinguisher rd 65500:2604
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-target export-rt target:65500:2604
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-target import-rt target:65500:2604
+set / network-instance peeringlan bridge-table proxy-arp admin-state enable
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning admin-state enable
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning age-time 600
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning send-refresh 180
+set / routing-policy policy ospf statement 100 match protocol host
+set / routing-policy policy ospf statement 100 action policy-result accept
+set / routing-policy policy ospf statement 200 match protocol ospfv2
+set / routing-policy policy ospf statement 200 action policy-result accept
+set / tunnel-interface vxlan1 vxlan-interface 2604 type bridged
+set / tunnel-interface vxlan1 vxlan-interface 2604 ingress vni 2604
+set / tunnel-interface vxlan1 vxlan-interface 2604 egress source-ip use-system-ipv4-address

static/assets/frys-ix/nikhef.conf

@@ -0,0 +1,132 @@
+set / interface ethernet-1/1 admin-state enable
+set / interface ethernet-1/1 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/1 subinterface 0 type routed
+set / interface ethernet-1/1 subinterface 0 admin-state enable
+set / interface ethernet-1/1 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/1 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/1 subinterface 0 ipv4 address 198.19.17.2/31
+set / interface ethernet-1/1 subinterface 0 ipv6 admin-state enable
+set / interface ethernet-1/2 admin-state enable
+set / interface ethernet-1/2 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/2 subinterface 0 type routed
+set / interface ethernet-1/2 subinterface 0 admin-state enable
+set / interface ethernet-1/2 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/2 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/2 subinterface 0 ipv4 address 198.19.17.4/31
+set / interface ethernet-1/2 subinterface 0 ipv6 admin-state enable
+set / interface ethernet-1/3 admin-state enable
+set / interface ethernet-1/3 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/3 subinterface 0 type routed
+set / interface ethernet-1/3 subinterface 0 admin-state enable
+set / interface ethernet-1/3 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/3 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/3 subinterface 0 ipv4 address 198.19.17.6/31
+set / interface ethernet-1/3 subinterface 0 ipv6 admin-state enable
+set / interface ethernet-1/4 admin-state enable
+set / interface ethernet-1/4 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/4 subinterface 0 type routed
+set / interface ethernet-1/4 subinterface 0 admin-state enable
+set / interface ethernet-1/4 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/4 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/4 subinterface 0 ipv4 address 198.19.17.8/31
+set / interface ethernet-1/4 subinterface 0 ipv6 admin-state enable
+set / interface ethernet-1/9 admin-state enable
+set / interface ethernet-1/9 breakout-mode num-breakout-ports 4
+set / interface ethernet-1/9 breakout-mode breakout-port-speed 10G
+set / interface ethernet-1/9/1 admin-state disable
+set / interface ethernet-1/9/2 admin-state disable
+set / interface ethernet-1/9/3 admin-state enable
+set / interface ethernet-1/9/3 vlan-tagging true
+set / interface ethernet-1/9/3 subinterface 0 type bridged
+set / interface ethernet-1/9/3 subinterface 0 admin-state enable
+set / interface ethernet-1/9/3 subinterface 0 vlan encap untagged
+set / interface ethernet-1/9/4 admin-state disable
+set / interface ethernet-1/29 admin-state enable
+set / interface ethernet-1/29 subinterface 0 type routed
+set / interface ethernet-1/29 subinterface 0 admin-state enable
+set / interface ethernet-1/29 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/29 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/29 subinterface 0 ipv4 address 198.19.17.1/31
+set / interface ethernet-1/29 subinterface 0 ipv6 admin-state enable
+set / interface lo0 admin-state enable
+set / interface lo0 subinterface 0 admin-state enable
+set / interface lo0 subinterface 0 ipv4 admin-state enable
+set / interface lo0 subinterface 0 ipv4 address 198.19.16.1/32
+set / interface mgmt0 admin-state enable
+set / interface mgmt0 subinterface 0 admin-state enable
+set / interface mgmt0 subinterface 0 ipv4 admin-state enable
+set / interface mgmt0 subinterface 0 ipv4 dhcp-client
+set / interface mgmt0 subinterface 0 ipv6 admin-state enable
+set / interface mgmt0 subinterface 0 ipv6 dhcp-client
+set / interface system0 admin-state enable
+set / interface system0 subinterface 0 admin-state enable
+set / interface system0 subinterface 0 ipv4 admin-state enable
+set / interface system0 subinterface 0 ipv4 address 198.19.18.1/32
+set / network-instance default type default
+set / network-instance default admin-state enable
+set / network-instance default description "fabric: dc1 role: spine"
+set / network-instance default router-id 198.19.16.1
+set / network-instance default ip-forwarding receive-ipv4-check false
+set / network-instance default interface ethernet-1/1.0
+set / network-instance default interface ethernet-1/2.0
+set / network-instance default interface ethernet-1/29.0
+set / network-instance default interface ethernet-1/3.0
+set / network-instance default interface ethernet-1/4.0
+set / network-instance default interface lo0.0
+set / network-instance default interface system0.0
+set / network-instance default protocols bgp admin-state enable
+set / network-instance default protocols bgp autonomous-system 65500
+set / network-instance default protocols bgp router-id 198.19.16.1
+set / network-instance default protocols bgp dynamic-neighbors accept match 198.19.16.0/24 peer-group overlay
+set / network-instance default protocols bgp afi-safi evpn admin-state enable
+set / network-instance default protocols bgp preference ibgp 170
+set / network-instance default protocols bgp route-advertisement rapid-withdrawal true
+set / network-instance default protocols bgp route-advertisement wait-for-fib-install false
+set / network-instance default protocols bgp group overlay peer-as 65500
+set / network-instance default protocols bgp group overlay afi-safi evpn admin-state enable
+set / network-instance default protocols bgp group overlay afi-safi ipv4-unicast admin-state disable
+set / network-instance default protocols bgp group overlay local-as as-number 65500
+set / network-instance default protocols bgp group overlay route-reflector client true
+set / network-instance default protocols bgp group overlay transport local-address 198.19.16.1
+set / network-instance default protocols bgp neighbor 198.19.16.0 admin-state enable
+set / network-instance default protocols bgp neighbor 198.19.16.0 peer-group overlay
+set / network-instance default protocols ospf instance default admin-state enable
+set / network-instance default protocols ospf instance default version ospf-v2
+set / network-instance default protocols ospf instance default router-id 198.19.16.1
+set / network-instance default protocols ospf instance default export-policy ospf
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/1.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/2.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/3.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/4.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/29.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface lo0.0 passive true
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface system0.0
+set / network-instance mgmt type ip-vrf
+set / network-instance mgmt admin-state enable
+set / network-instance mgmt description "Management network instance"
+set / network-instance mgmt interface mgmt0.0
+set / network-instance mgmt protocols linux import-routes true
+set / network-instance mgmt protocols linux export-routes true
+set / network-instance mgmt protocols linux export-neighbors true
+set / network-instance peeringlan type mac-vrf
+set / network-instance peeringlan admin-state enable
+set / network-instance peeringlan interface ethernet-1/9/3.0
+set / network-instance peeringlan vxlan-interface vxlan1.2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 admin-state enable
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 vxlan-interface vxlan1.2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 evi 2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 routes bridge-table mac-ip advertise true
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-distinguisher rd 65500:2604
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-target export-rt target:65500:2604
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-target import-rt target:65500:2604
+set / network-instance peeringlan bridge-table proxy-arp admin-state enable
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning admin-state enable
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning age-time 600
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning send-refresh 180
+set / routing-policy policy ospf statement 100 match protocol host
+set / routing-policy policy ospf statement 100 action policy-result accept
+set / routing-policy policy ospf statement 200 match protocol ospfv2
+set / routing-policy policy ospf statement 200 action policy-result accept
+set / tunnel-interface vxlan1 vxlan-interface 2604 type bridged
+set / tunnel-interface vxlan1 vxlan-interface 2604 ingress vni 2604
+set / tunnel-interface vxlan1 vxlan-interface 2604 egress source-ip use-system-ipv4-address

static/assets/frys-ix/nokia-leaf.conf

@@ -0,0 +1,105 @@
+set / interface ethernet-1/9 admin-state enable
+set / interface ethernet-1/9 vlan-tagging true
+set / interface ethernet-1/9 ethernet port-speed 10G
+set / interface ethernet-1/9 subinterface 0 type bridged
+set / interface ethernet-1/9 subinterface 0 admin-state enable
+set / interface ethernet-1/9 subinterface 0 vlan encap untagged
+set / interface ethernet-1/53 admin-state enable
+set / interface ethernet-1/53 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/53 subinterface 0 admin-state enable
+set / interface ethernet-1/53 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/53 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/53 subinterface 0 ipv4 address 198.19.17.11/31
+set / interface ethernet-1/53 subinterface 0 ipv6 admin-state enable
+set / interface ethernet-1/55 admin-state enable
+set / interface ethernet-1/55 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/55 subinterface 0 admin-state enable
+set / interface ethernet-1/55 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/55 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/55 subinterface 0 ipv4 address 198.19.17.7/31
+set / interface ethernet-1/55 subinterface 0 ipv6 admin-state enable
+set / interface ethernet-1/56 admin-state enable
+set / interface ethernet-1/56 ethernet forward-error-correction fec-option rs-528
+set / interface ethernet-1/56 subinterface 0 admin-state enable
+set / interface ethernet-1/56 subinterface 0 ip-mtu 9190
+set / interface ethernet-1/56 subinterface 0 ipv4 admin-state enable
+set / interface ethernet-1/56 subinterface 0 ipv4 address 198.19.17.9/31
+set / interface ethernet-1/56 subinterface 0 ipv6 admin-state enable
+set / interface lo0 admin-state enable
+set / interface lo0 subinterface 0 admin-state enable
+set / interface lo0 subinterface 0 ipv4 admin-state enable
+set / interface lo0 subinterface 0 ipv4 address 198.19.16.3/32
+set / interface mgmt0 admin-state enable
+set / interface mgmt0 subinterface 0 admin-state enable
+set / interface mgmt0 subinterface 0 ipv4 admin-state enable
+set / interface mgmt0 subinterface 0 ipv4 dhcp-client
+set / interface mgmt0 subinterface 0 ipv6 admin-state enable
+set / interface mgmt0 subinterface 0 ipv6 dhcp-client
+set / interface system0 admin-state enable
+set / interface system0 subinterface 0 admin-state enable
+set / interface system0 subinterface 0 ipv4 admin-state enable
+set / interface system0 subinterface 0 ipv4 address 198.19.18.3/32
+set / network-instance default type default
+set / network-instance default admin-state enable
+set / network-instance default description "fabric: dc1 role: leaf"
+set / network-instance default router-id 198.19.16.3
+set / network-instance default ip-forwarding receive-ipv4-check false
+set / network-instance default interface ethernet-1/53.0
+set / network-instance default interface ethernet-1/55.0
+set / network-instance default interface ethernet-1/56.0
+set / network-instance default interface lo0.0
+set / network-instance default interface system0.0
+set / network-instance default protocols bgp admin-state enable
+set / network-instance default protocols bgp autonomous-system 65500
+set / network-instance default protocols bgp router-id 198.19.16.3
+set / network-instance default protocols bgp afi-safi evpn admin-state enable
+set / network-instance default protocols bgp preference ibgp 170
+set / network-instance default protocols bgp route-advertisement rapid-withdrawal true
+set / network-instance default protocols bgp route-advertisement wait-for-fib-install false
+set / network-instance default protocols bgp group overlay peer-as 65500
+set / network-instance default protocols bgp group overlay afi-safi evpn admin-state enable
+set / network-instance default protocols bgp group overlay afi-safi ipv4-unicast admin-state disable
+set / network-instance default protocols bgp group overlay local-as as-number 65500
+set / network-instance default protocols bgp group overlay transport local-address 198.19.16.3
+set / network-instance default protocols bgp neighbor 198.19.16.0 admin-state enable
+set / network-instance default protocols bgp neighbor 198.19.16.0 peer-group overlay
+set / network-instance default protocols bgp neighbor 198.19.16.1 admin-state enable
+set / network-instance default protocols bgp neighbor 198.19.16.1 peer-group overlay
+set / network-instance default protocols ospf instance default admin-state enable
+set / network-instance default protocols ospf instance default version ospf-v2
+set / network-instance default protocols ospf instance default router-id 198.19.16.3
+set / network-instance default protocols ospf instance default export-policy ospf
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/53.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/55.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface ethernet-1/56.0 interface-type point-to-point
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface lo0.0 passive true
+set / network-instance default protocols ospf instance default area 0.0.0.0 interface system0.0
+set / network-instance mgmt type ip-vrf
+set / network-instance mgmt admin-state enable
+set / network-instance mgmt description "Management network instance"
+set / network-instance mgmt interface mgmt0.0
+set / network-instance mgmt protocols linux import-routes true
+set / network-instance mgmt protocols linux export-routes true
+set / network-instance mgmt protocols linux export-neighbors true
+set / network-instance peeringlan type mac-vrf
+set / network-instance peeringlan admin-state enable
+set / network-instance peeringlan interface ethernet-1/9.0
+set / network-instance peeringlan vxlan-interface vxlan1.2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 admin-state enable
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 vxlan-interface vxlan1.2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 evi 2604
+set / network-instance peeringlan protocols bgp-evpn bgp-instance 1 routes bridge-table mac-ip advertise true
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-distinguisher rd 65500:2604
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-target export-rt target:65500:2604
+set / network-instance peeringlan protocols bgp-vpn bgp-instance 1 route-target import-rt target:65500:2604
+set / network-instance peeringlan bridge-table proxy-arp admin-state enable
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning admin-state enable
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning age-time 600
+set / network-instance peeringlan bridge-table proxy-arp dynamic-learning send-refresh 180
+set / routing-policy policy ospf statement 100 match protocol host
+set / routing-policy policy ospf statement 100 action policy-result accept
+set / routing-policy policy ospf statement 200 match protocol ospfv2
+set / routing-policy policy ospf statement 200 action policy-result accept
+set / tunnel-interface vxlan1 vxlan-interface 2604 type bridged
+set / tunnel-interface vxlan1 vxlan-interface 2604 ingress vni 2604
+set / tunnel-interface vxlan1 vxlan-interface 2604 egress source-ip use-system-ipv4-address

static/mta-sts.txt

@@ -0,0 +1,4 @@
+version: STSv1
+mode: enforce
+mx: smtp.ipng.ch
+max_age: 86400

static/prefixes.txt

@@ -0,0 +1,21 @@
+# Source: https://ipng.ch/sshkeys.txt
+#
+
+194.1.163.64/27                # AS8298 IPng Networks
+2001:678:d78:3::/64            # AS8298 IPng Networks 
+94.142.241.184/29              # AS8283 Coloclue
+2a02:898:146::/64              # AS8283 Coloclue
+46.20.246.112/28               # AS25091 IP-Max
+2a02:2528:ff00::/64            # AS25091 IP-Max
+193.109.122.0/26               # AS12859 BIT
+2001:7b8:3:1e::/64             # AS12859 BIT
+
+2001:678:d78:300::/56          # IPng Wireguard
+46.20.243.179/32               # IPng Trusted border0.nlams3
+2a02:2528:ff02::179/128        # IPng Trusted border0.nlams3
+194.1.163.153/32               # IPng Trusted border0.chplo0
+2001:678:d78:7::2/128          # IPng Trusted border0.chplo0
+194.1.163.190/32               # IPng Trusted border0.chrma0
+2001:678:d78:b::2/128          # IPng Trusted border0.chrma0
+194.1.163.68/32                # IPng Trusted border0.chbtl0
+2001:678:d78:3::4/128          # IPng Trusted border0.chbtl0

static/sshkeys.txt

@@ -0,0 +1,11 @@
+# Source: https://ipng.ch/sshkeys.txt
+#
+# OpenBSD bastion
+ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIAXMfDOJtI3JztcPJ1DZMXzILZzMilMvodvMIfqqa1qr pim+openbsd@ipng.ch
+
+# Mac Studio (Secretive)
+ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBMtJZgTDWxBEbQ2vPYtOw4L0s4VRKUUjpu6aFPVx3CpqrjLpyJIxzBWTfb/VnEp95VfgM8IUAYYM8w7xoLd7QZc= pim+jessica+secretive@ipng.ch
+
+# Macbook Air M4 (Secretive)
+ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBASymGKXfKkfsYbo7UDrIBxl1F6X7LmVPQ3XOFOKp8tLI6zLyCYs5zgRNs/qksHOgKUK+fE/TzJ4XJsuSbYNMB0= pim+tammy+secretive@ipng.ch
+

themes/hugo-theme-ipng/assets/styles.scss

@@ -17,6 +17,7 @@ $text-very-light: #767676;
 $medium-light-text: #4f4a5f;
 $code-background: #f3f3f3;
 $codeblock-background: #f6f8fa;
+$codeblock-text: #99a;
 $code-text: #f8f8f2;
 $ipng-orange: #f46524;
 $ipng-darkorange: #8c1919;
@@ -63,7 +64,7 @@ body {
     line-height: $base-line-height * 1.0;
     font-family: BlinkMacSystemFont, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol";
 
-    @media #{$mq-mini} { font-size: $base-font-size * 0.8; }
+    /* $mq-mini or smaller */ font-size: $base-font-size * 0.8;
     @media #{$mq-small} { font-size: $base-font-size * 0.9; }
     @media #{$mq-medium} { font-size: $base-font-size * 1.0; }
     @media #{$mq-large} { font-size: $base-font-size * 1.0; }
@@ -78,8 +79,8 @@ main {
     padding: 1em 1em;
     padding-bottom: 5em;
 
-    @media #{$mq-mini} { margin: 0 2%; }
-    @media #{$mq-small} { margin: 0 2%; }
+    /* $mq-mini or smaller */ margin: 0 2%;
+    @media #{$mq-small} { margin: 0 5%; }
     @media #{$mq-medium} { margin: 0 17%; }
     @media #{$mq-large} { margin: 0 21%; }
     @media #{$mq-xlarge} { margin: 0 24%; }
@@ -142,7 +143,7 @@ pre {
 
     code {
         background-color: transparent;
-        color: #444;
+        color: $codeblock-text;
     }
 }
 
@@ -341,7 +342,7 @@ nav li:hover {
     display: flex;
     flex-flow: row wrap;
 
-    @media #{$mq-mini} { margin: 0; width: 100%; font-size: $base-font-size * 0.8; }
+    /* $mq-mini or smaller: */ margin: 0; width: 100%; font-size: $base-font-size * 0.8;
     @media #{$mq-small} { margin: 0 5%; width: 90%; font-size: $base-font-size * 0.9; }
     @media #{$mq-medium} { margin: 0 17%; width: 66%; font-size: $base-font-size * 1.0; }
     @media #{$mq-large} { margin: 0 21%; width: 58%; font-size: $base-font-size * 1.0; }
@@ -367,7 +368,7 @@ nav li:hover {
     color: $text-light;
     background-color: #f7f7f7;
 
-    @media #{$mq-mini} { margin: 0; width: 100%; font-size: $base-font-size * 0.8; }
+    /* $mq-mini or smaller: */ margin: 0; width: 100%; font-size: $base-font-size * 0.8;
     @media #{$mq-small} { margin: 0 5%; width: 90%; font-size: $base-font-size * 0.9; }
     @media #{$mq-medium} { margin: 0 17%; width: 66%; font-size: $base-font-size * 1.0; }
     @media #{$mq-large} { margin: 0 21%; width: 58%; font-size: $base-font-size * 1.0; }

themes/hugo-theme-ipng/layouts/_default/baseof.html

@@ -1,5 +1,5 @@
 <!DOCTYPE html>
-<html>
+<html lang="en">
     {{- partial "head.html" . -}}
     <body>
         {{- partial "header.html" . -}}

themes/hugo-theme-ipng/layouts/partials/head.html

@@ -1,26 +1,26 @@
 
 <head>
     <title>{{ .Site.Title }} {{ with .Title }}- {{ . }} {{ end }}</title>
-    <link rel="stylesheet" type="text/css" href="{{ "css/fonts.css" | relURL }}" />
-    <link rel="stylesheet" type="text/css" href="{{ "css/fontawesome.css" | relURL }}" />
+    <link rel="stylesheet" type="text/css" href="{{ "css/fonts.css" | relURL }}">
+    <link rel="stylesheet" type="text/css" href="{{ "css/fontawesome.css" | relURL }}">
     {{ $options := dict "transpiler" "libsass" "targetPath" "css/styles.css" -}}
     {{ $style := resources.Get "styles.scss" | toCSS $options | minify | fingerprint -}}
     <link rel="stylesheet" type="text/css" href="{{ $style.RelPermalink }}">
     {{ with resources.Get "css/userstyles.css" }}
     <link rel="stylesheet" type="text/css" href="{{ .Permalink }}">
     {{ end -}}
-    <link rel="icon" href="/assets/logo/favicon/favicon.ico" type="image/x-icon" sizes="any" />
-    <link rel="apple-touch-icon" href="/assets/logo/favicon/apple-touch-icon.png" />
-    <link rel="manifest" href="/assets/logo/favicon/icon.manifest" />
+    <link rel="icon" href="/assets/logo/favicon/favicon.ico" type="image/x-icon" sizes="any">
+    <link rel="apple-touch-icon" href="/assets/logo/favicon/apple-touch-icon.png">
+    <link rel="manifest" href="/assets/logo/favicon/icon.manifest">
     <meta charset="UTF-8">
     <meta name="author" content="{{ .Site.Params.Author }}">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    {{ range .AlternativeOutputFormats -}}
-        {{ printf `<link rel="%s" type="%s" href="%s" title="%s" />` .Rel .MediaType.Type .Permalink $.Site.Title | safeHTML }}
-    {{ end -}}
+    {{- range .AlternativeOutputFormats }}
+    {{ printf `<link rel="%s" type="%s" href="%s" title="%s">` .Rel .MediaType.Type .Permalink $.Site.Title | safeHTML }}
+    {{- end }}
     <script defer data-domain="ipng.ch" data-api="/api/event" src="/js/script.js"></script>
-    {{ if eq .Params.asciinema true -}}
-    <link rel="stylesheet" type="text/css" href="{{ "css/asciinema-player.css" | relURL }}" />
+    {{- if eq .Params.asciinema true }}
+    <link rel="stylesheet" type="text/css" href="{{ "css/asciinema-player.css" | relURL }}">
     <script src="{{ "js/asciinema-player.min.js" | relURL }}"></script>
     {{- end }}
 </head>

themes/hugo-theme-ipng/layouts/shortcodes/image.html

@@ -10,7 +10,7 @@
     <img src="{{ .Get "src" | relURL }}"
          {{- if or (.Get "alt") (.Get "caption") }} alt="{{ with .Get "alt" }}{{ replace . "'" "&#39;" }}{{ else }}{{ .Get "caption" | markdownify| plainify }}{{ end }}"
          {{- end -}}
-    /> <!-- Closing img tag -->
+    > <!-- Closing img tag -->
     {{- if .Get "link" }}</a>{{ end -}}
     {{- if or (or (.Get "title") (.Get "caption")) (.Get "attr") }}
     <figcaption>