Pools, CLI, versioning, Debian packaging, HTTPS fix

- Replaced flat `backends: [...]` list on frontends with an ordered `pools:`
  list; each pool has a name and a map of backends with per-pool weights (0–100,
  default 100). Pools express priority: first pool with a healthy backend wins.
- Removed global backend weight (was on the backend, now lives in the pool).
- Config validation enforces non-empty pools, non-empty pool names, weight
  range, and consistent address families across all pools of a frontend.

- Added `PoolBackendInfo { name, weight }` and changed `PoolInfo.backends` from
  `repeated string` to `repeated PoolBackendInfo` so weights are visible over
  the API.

- Full interactive shell with readline, tab completion, and `?` inline help.
- Command tree parser (Walk) handles fixed keywords and dynamic slot nodes;
  prefix matching with exact-match priority.
- Commands: `show version/frontends/frontend/backends/backend/healthchecks/
  healthcheck`, `set backend <name> pause|resume`, `quit`/`exit`.
- `show frontend` output is hierarchical (pools → backends) with per-backend
  weights and `[disabled]` notation; pool section uses fixed-width formatting
  so ANSI color codes don't corrupt tabwriter alignment.
- `-color` flag (default true) wraps static field labels in dark-blue ANSI;
  works correctly with tabwriter because all labels carry identical-length
  escape sequences.

- `cmd/version.go` package holds `version`, `commit`, `date` vars set at build
  time via `-ldflags -X`.
- `make build` / `make build-amd64` / `make build-arm64` all inject
  `VERSION=0.1.1`, `COMMIT_HASH` (from `git rev-parse --short HEAD`), and
  `DATE` (UTC ISO-8601).
- `maglevc` prints version on interactive startup and exposes `show version`.
- `maglevd` logs version/commit/date at startup; `-version` flag prints and exits.

- `doHTTPProbe` was building a `https://` target URL even though TLS was already
  applied to the connection inside `inNetns`. `http.Transport` then wrapped the
  connection in a second TLS layer, producing "http: server gave HTTP response
  to HTTPS client". Fixed by always using `http://` in the target URL.
- Added `TestHTTPSProbe` using `httptest.NewTLSServer` to cover the full path.

- New `docs/user-guide.md`: maglevd flags/signals, maglevc commands, shell
  completion, and command-tree parser walkthrough.
- New `docs/healthchecks.md`: state machine, rise/fall model, probe intervals,
  all transition events with log examples.
- Updated `docs/config-guide.md`: pools design, removed global weight from
  backends, updated all examples.
- Updated `README.md`: packaging table, build paths, corrected binary locations
  (`/usr/sbin/maglevd`), config filename (`.yaml`).

- `debian/` directory contains `control.in`, `maglevd.service`, `default.maglev`,
  `maglev.yaml` (example config), `conffiles`, `postinst`, `prerm`.
- `debian/build-deb.sh` stages a package tree and calls `dpkg-deb`; emits
  `build/vpp-maglev_<version>~<commit>_<arch>.deb`.
- Cross-compiles for amd64 and arm64 in one `make pkg-deb` invocation.
- `maglevd` installed to `/usr/sbin/`, `maglevc` to `/usr/bin/`.
- Service reads `MAGLEV_CONFIG` from `/etc/default/maglev`
  (default: `/etc/maglev/maglev.yaml`).
- Man pages `maglevd(8)` and `maglevc(1)` live in `docs/` and are gzip'd into
  the package.
- All build output goes to `build/<arch>/`; `build/` is gitignored.
This commit is contained in:
2026-04-11 12:18:17 +02:00
parent ad7d7e20fc
commit d612086a5f
31 changed files with 1471 additions and 282 deletions

133
docs/user-guide.md Normal file
View File

@@ -0,0 +1,133 @@
# User Guide
## maglevd
`maglevd` is the health-checker daemon. It probes backends according to the
configuration file, maintains their health state, and exposes a gRPC API for
inspection and control.
### Flags
| Flag | Environment variable | Default | Description |
|---|---|---|---|
| `--config` | `MAGLEV_CONFIG` | `/etc/maglev/maglev.yaml` | Path to the YAML configuration file. |
| `--grpc-addr` | `MAGLEV_GRPC_ADDR` | `:9090` | TCP address on which the gRPC server listens. |
| `--log-level` | `MAGLEV_LOG_LEVEL` | `info` | Log verbosity: `debug`, `info`, `warn`, or `error`. |
| `--version` | — | — | Print version, commit hash, and build date, then exit. |
Flags take precedence over environment variables. Both are optional; defaults
are used for anything not set.
### Signals
| Signal | Effect |
|---|---|
| `SIGHUP` | Reload the configuration file. New backends are started, removed backends are stopped, backends whose health-check config is unchanged continue probing without interruption. |
| `SIGTERM` / `SIGINT` | Graceful shutdown. Active gRPC streams are closed, the server drains, then the process exits. |
### Capabilities
`maglevd` requires `CAP_NET_RAW` when any health check uses `type: icmp`.
All other check types (`tcp`, `http`) use normal TCP sockets and require no
special capabilities.
### Logging
All log output is written to stdout as JSON using Go's `log/slog`. The first
line logged after the logger is configured is a `starting` record that includes
`version`, `commit`, and `date`. Every state change emits a `backend-transition`
line at `INFO` level. Set `--log-level debug` to see individual probe attempts
and their outcomes.
---
## maglevc
`maglevc` is the interactive control-plane client. It connects to a running
`maglevd` over gRPC and either executes a single command or drops into an
interactive shell.
### Usage
```sh
maglevc [--server host:port] [--color[=bool]] [command...]
```
| Flag | Default | Description |
|---|---|---|
| `--server` | `localhost:9090` | Address of the `maglevd` gRPC server. |
| `--color` | `true` | Colorize static field labels in output (dark blue ANSI). Pass `--color=false` to disable, e.g. when piping. |
When `command` arguments are supplied the command is executed and `maglevc`
exits. When no arguments are given an interactive shell is started and the
build version is printed on entry.
### Commands
```
show version Print build version, commit hash, and build date.
show frontends List all frontend names.
show frontend <name> Show address, protocol, port, description, and pools.
Each pool lists its backends with weights (if != 100)
and marks disabled backends with [disabled].
show backends List all backend names.
show backend <name> Show address, current state (with duration in that state),
enabled flag, health check, and recent state transitions
with timestamps and how long ago each occurred.
show healthchecks List all health-check names.
show healthcheck <name> Show full health-check configuration.
set backend <name> pause Suspend health checking for a backend, freezing its state.
set backend <name> resume Resume health checking; backend re-enters unknown state
and is probed immediately.
quit / exit Leave the interactive shell.
```
### Interactive shell
The shell prompt is `maglev> `. Two completion mechanisms are available:
**Tab completion** — pressing `<Tab>` at any point completes the current token.
Fixed keywords (commands and subcommands) are completed from the command tree.
Backend, frontend, and health-check names are fetched live from the server with
a 1-second timeout. If the partial token is unambiguous the word is completed
in place; if multiple candidates exist they are listed and the prompt is
restored.
**Inline help (`?`)** — typing `?` at any point prints the available
completions for the current position, with a short description next to each
keyword. The `?` character is not added to the input line.
Commands and keywords support **prefix matching**: typing `sh b` is equivalent
to `show backend` provided the prefix is unambiguous. Exact matches always take
priority over prefix matches, so `show backend` and `show backends` are
unambiguous even though one is a prefix of the other.
### Command tree and parser
Commands form a tree of `Node` values. Each node has a fixed `Word` (a keyword)
or is a *slot node* (marked by a `Dynamic` function that enumerates valid
values at completion time). The parser (`Walk`) descends the tree token by
token:
1. Try to match the current token against the fixed-keyword children of the
current node (exact match first, then unique prefix match).
2. If no fixed child matches, try a slot child — any token is accepted and
stored as an argument.
3. Stop when tokens are exhausted or no match is found.
The leaf node reached by `Walk` must have a `Run` function; otherwise the
available sub-commands at that position are printed as help. Arguments
collected from slot nodes are passed to `Run` as a slice.
Example walk for `set backend nginx0-ams pause`:
```
root → set → backend → <name>(nginx0-ams collected as arg) → pause
```
`pause.Run` is called with `args = ["nginx0-ams"]`.