Files
vpp-maglev/proto/maglev.proto
Pim van Pelt 224167ce39 Dataplane reconcile fixes; LB counters cleanup; SPA scope cookie
Checker / reload:
- Reload's update-in-place branch now mirrors b.Address onto the
  runtime health.Backend. Without this, GetBackend kept returning
  the pre-reload address indefinitely after a config edit that
  touched addresses but not healthcheck settings — the VPP sync
  path reads cfg.Backends directly so the dataplane moved on
  while the gRPC and SPA view stayed wedged on the old IPv4/IPv6.

Sync (internal/vpp/lbsync.go):
- reconcileVIP now detects encap mismatch in addition to
  src-ip-sticky mismatch and takes the full tear-down / re-add
  path via a new shared recreateVIP helper. Triggered when every
  backend flips address family (gre4 <-> gre6) and the existing
  VIP can no longer accept new ASes — previously the sync wedged
  with 'Invalid address family' until a full maglevd restart.
- setASWeight is issued whenever the state machine requests
  flush (a.Flush=true), not only on the weight-value transition
  edge. Fixes the case where a backend reached StateDisabled
  after its effective weight had already been drained to 0 by
  pool failover — the sticky-cache entries pointing at it were
  previously never cleared.

maglev-frontend:
- signal.Ignore(SIGHUP) so a controlling-terminal disconnect
  doesn't kill the daemon.
- debian/vpp-maglev.service grants CAP_SYS_ADMIN in addition to
  CAP_NET_RAW so setns(CLONE_NEWNET) can join the healthcheck
  netns. Comment documents the 'operation not permitted' symptom
  and notes the knob can be dropped if the deployment doesn't use
  the 'netns:' healthcheck option.

LB plugin counters (internal/vpp/lbstats.go + friends):
- Fix the VIP counter regex: the LB plugin registers
  vlib_simple_counter_main_t names without a leading '/'
  (vlib_validate_simple_counter in counter.c:50 uses cm->name
  verbatim; only entries that set cm->stat_segment_name get a
  slash). first/next/untracked/no-server now read through as
  live values instead of zero.
- Drop the per-backend FIB counter block end-to-end (proto,
  grpcapi, metrics, vpp.Client, lbstats, maglevc). Traced from
  lb/node.c:558 into ip{4,6}_forward.h:141 — the LB plugin
  forwards by writing adj_index[VLIB_TX] directly and bypassing
  ip{4,6}_lookup_inline, which is the only path that increments
  lbm_to_counters. The backend's FIB load_balance stats_index
  literally never ticks for LB-forwarded traffic, so the column
  was always zero and misleading. docs/implementation/TODO
  records the full investigation and the recommended upstream
  path (new lb_as_stats_dump API message) for when we're ready
  to carry that VPP patch.
- maglevc show vpp lb counters: plain-text tabular headers.
  label() wraps strings in ANSI escapes (~11 bytes of overhead),
  but tabwriter counts bytes, not rendered width — so a header
  row with label()'d cells and data rows with plain cells drifts
  column alignment on every row. color.go comment now spells
  out the constraint: label() only works when column N is
  wrapped identically in every row (key-value layouts are fine,
  multi-column tables with header-only labelling are not).

SPA:
- stores/scope.ts is cookie-backed (maglev_scope, 1 year,
  SameSite=Lax). App.tsx hydrates from the cookie then validates
  against the fetched snapshots: a cookie referencing a maglevd
  that no longer exists falls through to snaps[0] instead of
  leaving the user on a ghost selection.
- components/Flash.tsx wraps props.value in createMemo. Solid's
  on() fires its callback on every dep notification, not on
  value change — source is right in solid-js/dist/solid.js:460,
  no equality check. Without the memo, flipping scope between
  two 'connected' maglevds (or any other cross-store reactive
  re-eval that doesn't actually change the concrete string)
  replays the animation every time. createMemo's default ===
  dedupe fixes it in one place for every Flash consumer,
  superseding the local createMemo workaround we'd added in
  BackendRow earlier.
2026-04-14 14:40:16 +02:00

309 lines
9.6 KiB
Protocol Buffer

syntax = "proto3";
package maglev;
option go_package = "git.ipng.ch/ipng/vpp-maglev/internal/grpcapi";
// Maglev exposes the state of backend health for all frontends.
service Maglev {
rpc ListFrontends(ListFrontendsRequest) returns (ListFrontendsResponse);
rpc GetFrontend(GetFrontendRequest) returns (FrontendInfo);
rpc ListBackends(ListBackendsRequest) returns (ListBackendsResponse);
rpc GetBackend(GetBackendRequest) returns (BackendInfo);
rpc PauseBackend(BackendRequest) returns (BackendInfo);
rpc ResumeBackend(BackendRequest) returns (BackendInfo);
rpc EnableBackend(BackendRequest) returns (BackendInfo);
rpc DisableBackend(BackendRequest) returns (BackendInfo);
rpc ListHealthChecks(ListHealthChecksRequest) returns (ListHealthChecksResponse);
rpc GetHealthCheck(GetHealthCheckRequest) returns (HealthCheckInfo);
rpc SetFrontendPoolBackendWeight(SetWeightRequest) returns (FrontendInfo);
rpc WatchEvents(WatchRequest) returns (stream Event);
rpc CheckConfig(CheckConfigRequest) returns (CheckConfigResponse);
rpc ReloadConfig(ReloadConfigRequest) returns (ReloadConfigResponse);
rpc GetVPPInfo(GetVPPInfoRequest) returns (VPPInfo);
rpc GetVPPLBState(GetVPPLBStateRequest) returns (VPPLBState);
rpc SyncVPPLBState(SyncVPPLBStateRequest) returns (SyncVPPLBStateResponse);
rpc GetVPPLBCounters(GetVPPLBCountersRequest) returns (VPPLBCounters);
}
// ---- requests ---------------------------------------------------------------
message ListFrontendsRequest {}
message GetFrontendRequest {
string name = 1;
}
message ListBackendsRequest {}
message GetBackendRequest {
string name = 1;
}
message BackendRequest {
string name = 1;
}
message ListHealthChecksRequest {}
message GetHealthCheckRequest {
string name = 1;
}
message CheckConfigRequest {}
message CheckConfigResponse {
bool ok = 1;
string parse_error = 2; // set when YAML cannot be read or parsed
string semantic_error = 3; // set when YAML is valid but semantically incorrect
}
message ReloadConfigRequest {}
message ReloadConfigResponse {
bool ok = 1;
string parse_error = 2; // set when YAML cannot be read or parsed
string semantic_error = 3; // set when YAML is valid but semantically incorrect
string reload_error = 4; // set when config is valid but the reload itself failed
}
message GetVPPInfoRequest {}
message VPPInfo {
string version = 1;
string build_date = 2;
string build_directory = 3;
uint32 pid = 4;
int64 boottime_ns = 5; // unix timestamp (ns) when VPP started (from /sys/boottime)
int64 connecttime_ns = 6; // unix timestamp (ns) when maglevd connected to VPP
}
// ---- VPP load-balancer state ------------------------------------------------
message GetVPPLBStateRequest {}
// VPPLBConf mirrors VPP's lb_conf_get_reply: global LB plugin settings.
message VPPLBConf {
string ip4_src_address = 1;
string ip6_src_address = 2;
uint32 sticky_buckets_per_core = 3;
uint32 flow_timeout = 4;
}
// VPPLBAS is one application server attached to a VIP.
message VPPLBAS {
string address = 1;
uint32 weight = 2; // 0-100
uint32 flags = 3; // VPP AS flags (bit 0 = used, bit 1 = flushed)
uint32 num_buckets = 4;
int64 in_use_since_ns = 5; // unix timestamp (ns), 0 if never used
}
// VPPLBVIP mirrors VPP's lb_vip_details plus the attached application servers.
// Note: srv_type, dscp, and target_port are intentionally omitted — maglevd
// only supports GRE encap, so NAT/L3DSR-specific fields don't apply.
message VPPLBVIP {
string prefix = 1; // CIDR, e.g. 192.0.2.1/32
uint32 protocol = 2; // 6=TCP, 17=UDP, 255=any
uint32 port = 3; // 0 = all-port VIP
string encap = 4; // gre4|gre6|l3dsr|nat4|nat6
uint32 flow_table_length = 5;
repeated VPPLBAS application_servers = 6;
bool src_ip_sticky = 7; // source-IP based sticky session (scraped via cli_inband)
}
message VPPLBState {
VPPLBConf conf = 1;
repeated VPPLBVIP vips = 2;
}
// SyncVPPLBStateRequest triggers a reconciliation between the maglev config
// and the VPP load-balancer dataplane. When frontend_name is set, only that
// frontend's VIP is synced (SyncLBStateVIP) and no VIPs are removed. When
// unset, a full reconciliation runs (SyncLBStateAll), which will also remove
// stale VIPs from VPP.
message SyncVPPLBStateRequest {
optional string frontend_name = 1;
}
message SyncVPPLBStateResponse {}
// ---- VPP load-balancer runtime counters ------------------------------------
// GetVPPLBCountersRequest asks maglevd for the most recent per-VIP and
// per-backend counter snapshot. The data is served from an in-process
// cache that is refreshed every ~5 seconds server-side; the call itself
// is cheap and does not hit VPP.
message GetVPPLBCountersRequest {}
// VPPLBVIPCounters is the point-in-time counter row for a single VIP.
// The four lb_* fields are the LB plugin's SimpleCounters (packets only);
// packets / bytes come from the VPP FIB's combined counter at the VIP's
// host prefix (/net/route/to).
message VPPLBVIPCounters {
string prefix = 1; // CIDR, e.g. 192.0.2.1/32
string protocol = 2; // tcp | udp | any
uint32 port = 3;
uint64 next_packet = 4; // "/packet from existing sessions"
uint64 first_packet = 5; // "/first session packet"
uint64 untracked_packet = 6; // "/untracked packet"
uint64 no_server = 7; // "/no server configured"
uint64 packets = 8; // /net/route/to (FIB, summed across workers)
uint64 bytes = 9; // /net/route/to (FIB, summed across workers)
}
// VPPLBCounters wraps the per-VIP counter list returned by
// GetVPPLBCounters. There is no per-backend counter block: VPP's LB
// plugin forwarding node bypasses ip{4,6}_lookup_inline and writes
// adj_index[VLIB_TX] directly, so /net/route/to at a backend's FIB
// entry never ticks for LB-forwarded traffic — the four per-VIP
// counters are the only per-VIP-and-coarser signal VPP exposes today.
message VPPLBCounters {
repeated VPPLBVIPCounters vips = 1;
// Field 2 (repeated VPPLBBackendCounters backends) was removed; the
// index is reserved so a future replacement doesn't accidentally
// reuse it.
reserved 2;
}
message SetWeightRequest {
string frontend = 1;
string pool = 2;
string backend = 3;
int32 weight = 4; // 0-100
// flush, when true, also clears VPP's flow table for this backend
// so existing sessions are torn down. When false (default), only
// Maglev's new-bucket mapping is updated and live flows keep
// draining to this backend.
bool flush = 5;
}
// WatchRequest controls which event types are streamed. All fields default to
// true (i.e. an empty request subscribes to everything at info level).
message WatchRequest {
optional bool log = 1; // include log events (default: true)
string log_level = 2; // minimum log level: debug|info|warn|error (default: info)
optional bool backend = 3; // include backend transition events (default: true)
optional bool frontend = 4; // include frontend events (default: true)
}
// ---- responses --------------------------------------------------------------
message ListFrontendsResponse {
repeated string frontend_names = 1;
}
message PoolBackendInfo {
string name = 1;
int32 weight = 2; // configured weight from YAML (0-100)
int32 effective_weight = 3; // state-aware weight after pool-failover logic
}
message PoolInfo {
string name = 1;
repeated PoolBackendInfo backends = 2;
}
message FrontendInfo {
string name = 1;
string address = 2;
string protocol = 3;
uint32 port = 4;
repeated PoolInfo pools = 5;
string description = 6;
bool src_ip_sticky = 7; // VPP LB uses src-IP-based stickiness for this VIP
}
message ListBackendsResponse {
repeated string backend_names = 1;
}
message ListHealthChecksResponse {
repeated string names = 1;
}
message HTTPCheckParams {
string path = 1;
string host = 2;
int32 response_code_min = 3;
int32 response_code_max = 4;
string response_regexp = 5;
string server_name = 6;
bool insecure_skip_verify = 7;
}
message TCPCheckParams {
bool ssl = 1;
string server_name = 2;
bool insecure_skip_verify = 3;
}
message HealthCheckInfo {
string name = 1;
string type = 2;
uint32 port = 3;
string probe_ipv4_src = 4;
string probe_ipv6_src = 5;
int64 interval_ns = 6;
int64 fast_interval_ns = 7;
int64 down_interval_ns = 8;
int64 timeout_ns = 9;
int32 rise = 10;
int32 fall = 11;
HTTPCheckParams http = 12;
TCPCheckParams tcp = 13;
}
message BackendInfo {
string name = 1;
string address = 2;
string state = 3;
repeated TransitionRecord transitions = 4;
bool enabled = 5;
string healthcheck = 6;
}
message TransitionRecord {
string from = 1;
string to = 2;
int64 at_unix_ns = 3;
}
// ---- event stream -----------------------------------------------------------
// LogAttr is a single key/value attribute from a structured log record.
message LogAttr {
string key = 1;
string value = 2;
}
// LogEvent carries a single structured log record.
message LogEvent {
int64 at_unix_ns = 1;
string level = 2;
string msg = 3;
repeated LogAttr attrs = 4;
}
// BackendEvent is emitted on every backend state transition.
message BackendEvent {
string backend_name = 1;
TransitionRecord transition = 2;
}
// FrontendEvent is emitted when a frontend's aggregate state changes.
// Frontends have three states: unknown, up, down. See docs/healthchecks.md.
message FrontendEvent {
string frontend_name = 1;
TransitionRecord transition = 2;
}
// Event is the envelope returned by WatchEvents.
message Event {
oneof event {
LogEvent log = 1;
BackendEvent backend = 2;
FrontendEvent frontend = 3;
}
}