Failed to start daemon broker: ${lastError?.message ?? "sock

Error message

Failed to start daemon broker: ${lastError?.message ?? "socket unavailable"}

What it means

#connectOnce attempts to connect to the broker socket and, if unreachable, spawns the broker and retries within a deadline. When all attempts fail, it throws this aggregate error carrying the last underlying failure's message (or 'socket unavailable' if none was captured).

Source

Thrown at packages/coding-agent/src/launch/client.ts:305

			this.#bindSocket(await openSocket(this.#endpoint, 250));
			return;
		} catch {
			// No live broker. Multiple clients may race to spawn; the broker's PID
			// lease selects one winner before any candidate touches the socket.
		}
		this.#spawnBroker();
		const deadline = Date.now() + CONNECT_TIMEOUT_MS;
		let lastError: Error | undefined;
		while (Date.now() < deadline) {
			try {
				this.#bindSocket(await openSocket(this.#endpoint, 250));
				return;
			} catch (error) {
				lastError = error instanceof Error ? error : new Error(String(error));
				await Bun.sleep(CONNECT_RETRY_MS);
			}
		}
		throw new Error(`Failed to start daemon broker: ${lastError?.message ?? "socket unavailable"}`);
	}

	#spawnBroker(): void {
		const spawn = resolveWorkerSpawnCmd(DAEMON_BROKER_WORKER_ARG);
		const overlay: Record<string, string> = {
			[DAEMON_PROJECT_DIR_ENV]: this.projectDir,
			[DAEMON_RUNTIME_DIR_ENV]: this.#runtimeDir,
		};
		if (this.#idleGraceMs !== undefined) overlay[DAEMON_IDLE_GRACE_ENV] = String(this.#idleGraceMs);
		const child = Bun.spawn(spawn.cmd, {
			cwd: spawn.cwd,
			env: workerEnvFromParent(overlay),
			stdin: "ignore",
			stdout: "ignore",
			stderr: "ignore",
			...BROKER_SPAWN_OPTIONS,
		});
		child.unref();

View on GitHub (pinned to 9690622007)

Solutions

Read lastError in the message — fix the underlying broker startup failure first
Verify the runtime dir is writable and the worker entry resolves (workerHostEntry / fallback)
Run the broker in the foreground to see its real startup error, then retry the client

Example fix

// before
await client.request(op); // throws aggregate start failure
// after
try { await client.request(op); }
catch (err) {
  logger.error("daemon broker start failed", { cause: err.message }); // inspect lastError, fix root cause
  throw err;
}

Defensive patterns

Strategy: retry

Validate before calling

const socketPath = daemonBrokerEndpoint(projectDir, runtimeDir);
await fs.mkdir(runtimeDir, { recursive: true });
await fs.access(runtimeDir, fs.constants.W_OK); // runtime dir must be writable before connecting

Try / catch

try {
  await client.request({ op: 'ping' });
} catch (err) {
  if (err.message.startsWith('Failed to start daemon broker')) {
    logger.error('broker start failed', { cause: err.message }); // message embeds lastError — fix root cause
  }
  throw err;
}

Prevention

Verify runtimeDir is on a filesystem supporting unix sockets (not some network mounts)
Ensure the worker host entry resolves before spawning
Run the broker manually once to observe its real startup error

When it happens

Trigger: Broker repeatedly fails to start (bad env, missing worker entry, crash loop) or the socket path is unusable, exhausting the connect deadline.

Common situations: Runtime dir on a filesystem without unix socket support; broker worker crashing at startup (see env/token errors); permission issues on the socket file; port/path collisions between projects.

Related errors

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/fec571f43f55cc91. Report an issue: GitHub.