feat: hot-reload de agentes individuales via SIGHUP
Implementa el mecanismo de hot-reload descrito en el issue 0013: - agents/runtime.go: añade Agent.Stop() y Agent.Done() para ciclo de vida individual. Run() crea un contexto hijo cancelable y cierra el canal done al retornar. - cmd/launcher/registry.go (nuevo): agentRegistry rastrea agentes vivos por ID. Métodos: register, stopAndWait, reload, reloadAll, waitAll, cleanupLogs. reload() sigue el flujo completo: stop→wait→unsubscribe →reload config→recreate→rewire bus/orch→start nueva goroutine. - cmd/launcher/main.go: usa agentRegistry en lugar de sync.WaitGroup. Añade handler de SIGHUP en goroutine separada que lee run/reload.txt para determinar el agente objetivo (* o vacío = todos). Tras leer, borra run/reload.txt para no afectar el siguiente SIGHUP. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
+64
-28
@@ -12,7 +12,6 @@ import (
|
||||
"os"
|
||||
"os/signal"
|
||||
"path/filepath"
|
||||
"sync"
|
||||
"syscall"
|
||||
"time"
|
||||
|
||||
@@ -72,13 +71,7 @@ func main() {
|
||||
logger.Warn("could not create file logger, falling back to stdout", "err", err)
|
||||
launcherCleanup = func() {}
|
||||
}
|
||||
var cleanups []func()
|
||||
cleanups = append(cleanups, launcherCleanup)
|
||||
defer func() {
|
||||
for _, fn := range cleanups {
|
||||
fn()
|
||||
}
|
||||
}()
|
||||
defer launcherCleanup()
|
||||
|
||||
if len(configPaths) == 0 {
|
||||
logger.Warn("no agent configs found — nothing to start")
|
||||
@@ -96,15 +89,49 @@ func main() {
|
||||
if err != nil {
|
||||
// Non-fatal: orchestration is optional
|
||||
logger.Warn("orchestrator not started", "err", err)
|
||||
} else {
|
||||
} else if orch != nil {
|
||||
logger.Info("orchestrator initialized")
|
||||
}
|
||||
|
||||
// ── Start normal agents ──
|
||||
var wg sync.WaitGroup
|
||||
var scannerOnce sync.Once
|
||||
var scanner *mautrix.Client
|
||||
// ── Shared dependencies for agent registry ──
|
||||
deps := &launchDeps{
|
||||
agentBus: agentBus,
|
||||
orch: orch,
|
||||
logDir: logDir,
|
||||
logLevel: lvl,
|
||||
parentCtx: ctx,
|
||||
}
|
||||
registry := newAgentRegistry(deps)
|
||||
|
||||
// ── SIGHUP: hot-reload individual agent or all agents ──
|
||||
sighup := make(chan os.Signal, 1)
|
||||
signal.Notify(sighup, syscall.SIGHUP)
|
||||
go func() {
|
||||
for {
|
||||
select {
|
||||
case <-ctx.Done():
|
||||
return
|
||||
case _, ok := <-sighup:
|
||||
if !ok {
|
||||
return
|
||||
}
|
||||
id := readReloadTarget("run/reload.txt")
|
||||
// Remove the target file after reading so it doesn't
|
||||
// affect the next SIGHUP.
|
||||
_ = os.Remove("run/reload.txt")
|
||||
if id == "" {
|
||||
logger.Info("sighup: reloading all agents")
|
||||
registry.reloadAll(rulesFor)
|
||||
} else {
|
||||
logger.Info("sighup: reloading agent", "id", id)
|
||||
registry.reload(id, rulesFor)
|
||||
}
|
||||
}
|
||||
}
|
||||
}()
|
||||
|
||||
// ── Start normal agents ──
|
||||
var scannerOnce scanOnce
|
||||
for _, path := range configPaths {
|
||||
path := path
|
||||
cfg, err := config.Load(path)
|
||||
@@ -130,11 +157,11 @@ func main() {
|
||||
agentLogger = logger.With("agent", cfg.Agent.ID)
|
||||
agentCleanup = func() {}
|
||||
}
|
||||
cleanups = append(cleanups, agentCleanup)
|
||||
|
||||
a, err := agents.New(cfg, rules, agentLogger)
|
||||
if err != nil {
|
||||
logger.Error("failed to create agent", "id", cfg.Agent.ID, "err", err)
|
||||
agentCleanup()
|
||||
continue
|
||||
}
|
||||
|
||||
@@ -154,30 +181,28 @@ func main() {
|
||||
})
|
||||
|
||||
// Grab the first available Matrix client for room scanning
|
||||
scannerOnce.Do(func() {
|
||||
scanner = a.RawMatrixClient()
|
||||
})
|
||||
scannerOnce.set(a.RawMatrixClient())
|
||||
}
|
||||
|
||||
wg.Add(1)
|
||||
go func() {
|
||||
defer wg.Done()
|
||||
agentLogger.Info("agent running")
|
||||
if err := a.Run(ctx); err != nil {
|
||||
agentLogger.Error("agent stopped with error", "err", err)
|
||||
}
|
||||
}()
|
||||
registry.register(&runningAgent{
|
||||
agent: a,
|
||||
cfg: cfg,
|
||||
cfgPath: path,
|
||||
logger: agentLogger,
|
||||
logCleanup: agentCleanup,
|
||||
})
|
||||
}
|
||||
|
||||
// ── Startup room scan (after all participants are registered) ──
|
||||
if orch != nil && scanner != nil {
|
||||
orch.orchestrator.SetScanner(scanner)
|
||||
if orch != nil && scannerOnce.client != nil {
|
||||
orch.orchestrator.SetScanner(scannerOnce.client)
|
||||
scanCtx, scanCancel := context.WithTimeout(ctx, 30*time.Second)
|
||||
orch.orchestrator.ScanExistingRooms(scanCtx)
|
||||
scanCancel()
|
||||
}
|
||||
|
||||
wg.Wait()
|
||||
registry.waitAll()
|
||||
registry.cleanupLogs()
|
||||
logger.Info("all agents stopped")
|
||||
return nil
|
||||
},
|
||||
@@ -195,6 +220,17 @@ func main() {
|
||||
}
|
||||
}
|
||||
|
||||
// scanOnce captures the first Matrix client for room scanning.
|
||||
type scanOnce struct {
|
||||
client *mautrix.Client
|
||||
}
|
||||
|
||||
func (s *scanOnce) set(c *mautrix.Client) {
|
||||
if s.client == nil {
|
||||
s.client = c
|
||||
}
|
||||
}
|
||||
|
||||
// orchHandle wraps a running orchestrator with its config for the launcher.
|
||||
type orchHandle struct {
|
||||
orchestrator *orchshell.Orchestrator
|
||||
|
||||
Reference in New Issue
Block a user