Files
claudemesh/apps/cli/src/services/daemon/lifecycle.ts
Alejandro Gutiérrez 1a14cef1e0
Some checks failed
CI / Lint (push) Has been cancelled
CI / Typecheck (push) Has been cancelled
CI / Broker tests (Postgres) (push) Has been cancelled
CI / Docker build (linux/amd64) (push) Has been cancelled
feat(cli): 1.31.0 — session autoclean + broker verification + service path
Three operability fixes for users running the daemon under launchd or
systemd.

PID-watcher autoclean
=====================

The session reaper already dropped registry entries with dead pids on
a 30s loop, but had two real-world gaps:

- 30s sweep let stale presence linger on the broker for half a minute
- bare process.kill(pid, 0) trusts a recycled pid; a registry entry
  could survive its real owner's death whenever the OS rolled the
  pid number forward to a new program

Process-exit IPC from claude-code is best-effort and skipped on
SIGKILL / OOM / segfault / panic, so it cannot replace the sweep.

Fix:

- New process-info.ts captures opaque per-process start-times via
  ps -o lstart= (works on macOS and Linux, ~1 ms per call)
- registerSession stores the start-time alongside the pid
- reapDead drops entries when pid is dead OR start-time changed
  since register
- Sweep cadence 30s -> 5s
- Best-effort fallback to bare liveness when start-time capture
  fails at register time

Registry hooks already close the per-session broker WS on
deregister, so peer list rebuilds within one sweep of any session
exit.

Service-managed daemon: no more "spawn failed" false alarms
===========================================================

After claudemesh install (which writes a launchd plist or systemd
unit with KeepAlive=true), users routinely saw

  [claudemesh] warn daemon spawn failed: socket did not appear
  within 3000ms

even when the daemon was running fine. Two contributing causes:

1. Probe timeout was 800ms — the first IPC after a launchd-driven
   restart can take longer (SQLite migration + broker WS opens) and
   tripped it. Bumped to 2500ms.
2. On a failed probe the CLI tried its own detached spawn, which
   collided with launchd's KeepAlive restart cycle (singleton lock
   fails, child exits) and we'd then time out polling for a socket
   that was actually about to come up.

Now: when the launchd plist or systemd unit exists, the CLI does not
attempt a spawn. It waits up to 8s for the OS-managed unit to bring
the socket up. New service-not-ready state distinguishes "OS hasn't
restarted it yet" from "we tried to spawn and it failed".

Install verifies broker connectivity, not just process start
============================================================

Previously install ended once launchctl reported the unit loaded —
a daemon that boots but cannot reach the broker (blocked :443,
expired TLS, DNS, broker outage) only surfaced on the user's first
peer list or send.

/v1/health now includes per-mesh broker WS state. install polls it
for up to 15s after service boot and prints either "broker
connected (mesh=...)" or a warning naming the meshes still in
connecting state, with a hint at common causes.

The verification is best-effort and does not fail the install — it
just surfaces the issue early.

Tests
=====

4 new vitest cases cover the reaper paths: dead pid, live pid plus
matching start-time, live pid plus mismatched start-time (PID
reuse), and the no-start-time fallback. 83 of 83 pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-04 14:05:44 +01:00

299 lines
11 KiB
TypeScript

/**
* Daemon lifecycle helper — probe, auto-spawn, retry, fall-through.
*
* Every daemon-routed CLI verb passes through `ensureDaemonReady()` before
* its IPC call. The helper:
*
* 1. Probes the socket via a fast `/v1/version` IPC (~5-10 ms).
* 2. If the socket is missing OR present-but-stale, attempts a detached
* `claudemesh daemon up` spawn under a file-lock.
* 3. Polls for the new socket up to a budget (default 3s).
* 4. Returns a state describing what happened, so the caller can either
* proceed warm or fall back to the cold path with a clear warning.
*
* State machine:
* - "up" daemon was already running
* - "started" daemon was down; we spawned it; it came up
* - "down" daemon was down; auto-spawn skipped (e.g., recursion guard)
* - "spawn-failed" spawn attempted but socket never appeared within budget
* - "spawn-suppressed" recently-failed marker is fresh; skipped retry
*
* Stale-socket handling: if the socket file exists but the IPC probe
* fails (ECONNREFUSED / timeout), we treat the file as stale, remove
* it, and proceed as if the daemon were down. This fixes the prior bug
* where `existsSync(SOCK_FILE)` was a false positive after a daemon
* crash.
*
* Recursion guard: when we spawn the daemon we set
* `CLAUDEMESH_INTERNAL_NO_AUTOSPAWN=1` in its env so any nested CLI
* calls inside the daemon skip the auto-spawn check and avoid a loop.
*/
import { existsSync, readFileSync, statSync, unlinkSync, writeFileSync } from "node:fs";
import { homedir } from "node:os";
import { join } from "node:path";
import { ipc, IpcError } from "~/daemon/ipc/client.js";
import { DAEMON_PATHS } from "~/daemon/paths.js";
export type DaemonReadyState =
| "up"
| "started"
| "down"
| "spawn-failed"
| "spawn-suppressed"
/** 1.31.0+: launchd / systemd manages the daemon and it didn't respond
* within the service budget. Distinct from spawn-failed: the CLI did
* not attempt to spawn (the OS owns the lifecycle). */
| "service-not-ready";
export interface EnsureDaemonResult {
state: DaemonReadyState;
/** Total ms spent in this call (probe ± spawn ± poll). */
durationMs: number;
/** When state is `spawn-failed` or `spawn-suppressed`, a one-line reason. */
reason?: string;
}
export interface EnsureDaemonOpts {
/** Total budget for socket-appearance polling after spawn. Default 3000ms. */
budgetMs?: number;
/** Skip auto-spawn entirely. Used by `--no-daemon` and the recursion guard. */
noAutoSpawn?: boolean;
/** When auto-spawning a legacy single-mesh daemon, pin a slug. Omit for multi-mesh (default). */
mesh?: string;
}
const SPAWN_LOCK_FILE = () => join(DAEMON_PATHS.DAEMON_DIR, ".spawn.lock");
const SPAWN_FAIL_FILE = () => join(DAEMON_PATHS.DAEMON_DIR, ".spawn-failure");
const SPAWN_FAIL_TTL_MS = 30_000;
// 1.31.0: 800 ms was too tight — the daemon's first IPC after a launchd
// (re)start can take a beat while it migrates SQLite, opens broker WSes,
// and warms up the event loop. False "stale" probes triggered the
// pointless spawn → "socket did not appear" warning even on a perfectly
// healthy service-managed daemon. 2500 ms still bounds the worst case.
const PROBE_TIMEOUT_MS = 2_500;
// When the daemon is service-managed (launchd/systemd) and KeepAlive=true,
// the OS guarantees a restart on death — the CLI must NOT race that with
// its own spawn. Just wait longer for the service unit to come up.
const SERVICE_BUDGET_MS = 8_000;
let lastResultThisProcess: EnsureDaemonResult | null = null;
/** Probe daemon and return what we know. Cached per-process so a script
* with 50 sends doesn't re-spawn 50 times. */
export async function ensureDaemonReady(opts: EnsureDaemonOpts = {}): Promise<EnsureDaemonResult> {
if (lastResultThisProcess && (lastResultThisProcess.state === "up" || lastResultThisProcess.state === "started")) {
return lastResultThisProcess;
}
if (process.env.CLAUDEMESH_INTERNAL_NO_AUTOSPAWN === "1") {
opts = { ...opts, noAutoSpawn: true };
}
const result = await runEnsureDaemon(opts);
lastResultThisProcess = result;
return result;
}
/** Reset the per-process cache. Test helper. */
export function _resetDaemonReadyCache(): void {
lastResultThisProcess = null;
}
async function runEnsureDaemon(opts: EnsureDaemonOpts): Promise<EnsureDaemonResult> {
const t0 = Date.now();
// Step 1 — probe.
const probe = await probeDaemon();
if (probe === "up") return { state: "up", durationMs: Date.now() - t0 };
// Step 2 — service-managed shortcut. When launchd / systemd manages
// the daemon and KeepAlive is set, the OS will restart a crashed
// daemon on its own; the CLI must NOT race that with its own spawn
// (would double-bind the singleton lock and trigger "daemon already
// running" errors). Just wait quietly for the service to bring the
// socket up.
if (isServiceManaged()) {
if (probe === "stale") cleanupStaleFiles();
const polled = await pollForSocket(SERVICE_BUDGET_MS);
if (polled.ok) return { state: "up", durationMs: Date.now() - t0 };
const tool = process.platform === "darwin"
? `launchctl print gui/$(id -u)/${SERVICE_LABEL}`
: `systemctl --user status ${SYSTEMD_UNIT}`;
return {
state: "service-not-ready",
durationMs: Date.now() - t0,
reason: `service-managed daemon not responding within ${SERVICE_BUDGET_MS}ms (run \`${tool}\`)`,
};
}
if (probe === "stale") cleanupStaleFiles();
// Step 3 — auto-spawn unless forbidden.
if (opts.noAutoSpawn) {
return { state: "down", durationMs: Date.now() - t0, reason: "auto-spawn disabled" };
}
if (recentSpawnFailureFresh()) {
return {
state: "spawn-suppressed",
durationMs: Date.now() - t0,
reason: `daemon failed to start within last ${Math.round(SPAWN_FAIL_TTL_MS / 1000)}s`,
};
}
// Step 4 — spawn detached.
const spawnRes = await spawnDaemon(opts);
if (spawnRes.ok) {
return { state: "started", durationMs: Date.now() - t0 };
}
// Step 5 — record failure for backoff and report.
markSpawnFailure();
return { state: "spawn-failed", durationMs: Date.now() - t0, reason: spawnRes.reason };
}
const SERVICE_LABEL = "com.claudemesh.daemon";
const SYSTEMD_UNIT = "claudemesh-daemon.service";
/**
* Returns true when the user has installed the daemon as a launchd
* agent (macOS) or systemd --user unit (Linux). We detect by file
* presence rather than shelling out to launchctl/systemctl on every
* CLI invocation — this stays cheap and avoids spurious permission
* prompts on locked-down hosts.
*/
function isServiceManaged(): boolean {
if (process.platform === "darwin") {
return existsSync(join(homedir(), "Library", "LaunchAgents", `${SERVICE_LABEL}.plist`));
}
if (process.platform === "linux") {
return existsSync(join(homedir(), ".config", "systemd", "user", SYSTEMD_UNIT));
}
return false;
}
async function probeDaemon(): Promise<"up" | "absent" | "stale"> {
if (!existsSync(DAEMON_PATHS.SOCK_FILE)) return "absent";
try {
const res = await ipc<{ version?: string }>({ path: "/v1/version", timeoutMs: PROBE_TIMEOUT_MS });
if (res.status === 200) return "up";
return "stale";
} catch (err) {
if (err instanceof IpcError) return "stale";
const msg = String(err);
if (/ENOENT|ECONNREFUSED|ipc_timeout|EPIPE|ECONNRESET/.test(msg)) return "stale";
return "stale";
}
}
function cleanupStaleFiles(): void {
for (const p of [DAEMON_PATHS.SOCK_FILE, DAEMON_PATHS.PID_FILE]) {
try { unlinkSync(p); } catch { /* best-effort */ }
}
}
function recentSpawnFailureFresh(): boolean {
try {
const st = statSync(SPAWN_FAIL_FILE());
return Date.now() - st.mtimeMs < SPAWN_FAIL_TTL_MS;
} catch {
return false;
}
}
function markSpawnFailure(): void {
try { writeFileSync(SPAWN_FAIL_FILE(), String(Date.now()), { mode: 0o600 }); } catch { /* best-effort */ }
}
function clearSpawnFailure(): void {
try { unlinkSync(SPAWN_FAIL_FILE()); } catch { /* best-effort */ }
}
interface SpawnResult { ok: boolean; reason?: string; }
async function spawnDaemon(opts: EnsureDaemonOpts): Promise<SpawnResult> {
const lockResult = await acquireOrShareLock(opts);
if (lockResult === "wait-existing") {
// Another process is spawning; just wait for the socket to appear.
return await pollForSocket(opts.budgetMs ?? 3_000);
}
try {
const { spawn } = await import("node:child_process");
const binary = await resolveCliBinary();
const args = ["daemon", "up"];
if (opts.mesh) args.push("--mesh", opts.mesh);
const child = spawn(binary, args, {
detached: true,
stdio: "ignore",
env: { ...process.env, CLAUDEMESH_INTERNAL_NO_AUTOSPAWN: "1" },
});
child.unref();
const polled = await pollForSocket(opts.budgetMs ?? 3_000);
if (polled.ok) clearSpawnFailure();
return polled;
} catch (err) {
return { ok: false, reason: err instanceof Error ? err.message : String(err) };
} finally {
releaseLock();
}
}
/** Acquire spawn lock. If another process holds it AND its pid is alive,
* return "wait-existing" so we share that spawn attempt. If the pid is
* dead, take over the lock. */
async function acquireOrShareLock(_opts: EnsureDaemonOpts): Promise<"acquired" | "wait-existing"> {
const lockPath = SPAWN_LOCK_FILE();
if (existsSync(lockPath)) {
try {
const pidStr = readFileSync(lockPath, "utf8").trim();
const pid = Number.parseInt(pidStr, 10);
if (Number.isFinite(pid) && pid > 0) {
try {
process.kill(pid, 0); // signal 0 = liveness probe
return "wait-existing";
} catch {
// Holder is dead — fall through to take over.
}
}
} catch { /* unreadable lock — take over */ }
}
try {
writeFileSync(lockPath, String(process.pid), { mode: 0o600 });
} catch { /* best-effort; lock is advisory */ }
return "acquired";
}
function releaseLock(): void {
try { unlinkSync(SPAWN_LOCK_FILE()); } catch { /* best-effort */ }
}
async function pollForSocket(budgetMs: number): Promise<SpawnResult> {
const start = Date.now();
while (Date.now() - start < budgetMs) {
if (existsSync(DAEMON_PATHS.SOCK_FILE)) {
// Don't just trust file presence — confirm it answers.
const probe = await probeDaemon();
if (probe === "up") return { ok: true };
}
await new Promise((r) => setTimeout(r, 150));
}
return { ok: false, reason: `socket did not appear within ${budgetMs}ms` };
}
/** Resolve the absolute path to the `claudemesh` binary the user is running.
* When invoked via tsx/bun in dev, fall back to the system `claudemesh`. */
async function resolveCliBinary(): Promise<string> {
const argv1 = process.argv[1] ?? "claudemesh";
if (/\.ts$/.test(argv1) || /node_modules|src\/entrypoints/.test(argv1)) {
try {
const { execSync } = await import("node:child_process");
return execSync("which claudemesh", { encoding: "utf8" }).trim() || "claudemesh";
} catch {
return "claudemesh";
}
}
return argv1;
}