fix(server): use parent pid for lock ownership when spawned as child process
This commit is contained in:
@@ -0,0 +1,212 @@
|
|||||||
|
#!/usr/bin/env npx tsx
|
||||||
|
|
||||||
|
/**
|
||||||
|
* Regression: `paseo daemon stop` must stop supervised dev daemons
|
||||||
|
* without allowing daemon-runner to respawn a new worker process.
|
||||||
|
*/
|
||||||
|
|
||||||
|
import assert from 'node:assert'
|
||||||
|
import { spawn, spawnSync, type ChildProcess } from 'node:child_process'
|
||||||
|
import { mkdtemp, readFile, rm } from 'node:fs/promises'
|
||||||
|
import { tmpdir } from 'node:os'
|
||||||
|
import { join } from 'node:path'
|
||||||
|
import { $ } from 'zx'
|
||||||
|
|
||||||
|
$.verbose = false
|
||||||
|
|
||||||
|
const pollIntervalMs = 100
|
||||||
|
|
||||||
|
function sleep(ms: number): Promise<void> {
|
||||||
|
return new Promise((resolve) => setTimeout(resolve, ms))
|
||||||
|
}
|
||||||
|
|
||||||
|
function isProcessRunning(pid: number): boolean {
|
||||||
|
if (!Number.isInteger(pid) || pid <= 0) {
|
||||||
|
return false
|
||||||
|
}
|
||||||
|
|
||||||
|
try {
|
||||||
|
process.kill(pid, 0)
|
||||||
|
return true
|
||||||
|
} catch {
|
||||||
|
return false
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
type PidLockState = {
|
||||||
|
pid: number | null
|
||||||
|
}
|
||||||
|
|
||||||
|
async function readPidLockState(paseoHome: string): Promise<PidLockState> {
|
||||||
|
const pidPath = join(paseoHome, 'paseo.pid')
|
||||||
|
|
||||||
|
try {
|
||||||
|
const content = await readFile(pidPath, 'utf-8')
|
||||||
|
const parsed = JSON.parse(content) as { pid?: unknown }
|
||||||
|
const pid = typeof parsed.pid === 'number' && Number.isInteger(parsed.pid) && parsed.pid > 0
|
||||||
|
? parsed.pid
|
||||||
|
: null
|
||||||
|
return { pid }
|
||||||
|
} catch {
|
||||||
|
return { pid: null }
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
function readProcessCommand(pid: number): string | null {
|
||||||
|
const result = spawnSync('ps', ['-p', String(pid), '-o', 'command='], { encoding: 'utf8' })
|
||||||
|
if (result.status !== 0 || result.error) {
|
||||||
|
return null
|
||||||
|
}
|
||||||
|
const command = result.stdout.trim()
|
||||||
|
return command.length > 0 ? command : null
|
||||||
|
}
|
||||||
|
|
||||||
|
type DaemonStatus = {
|
||||||
|
status: string | null
|
||||||
|
pid: number | null
|
||||||
|
}
|
||||||
|
|
||||||
|
async function readDaemonStatus(paseoHome: string): Promise<DaemonStatus> {
|
||||||
|
const result =
|
||||||
|
await $`PASEO_HOME=${paseoHome} npx paseo daemon status --home ${paseoHome} --json`.nothrow()
|
||||||
|
if (result.exitCode !== 0) {
|
||||||
|
return { status: null, pid: null }
|
||||||
|
}
|
||||||
|
|
||||||
|
try {
|
||||||
|
const parsed = JSON.parse(result.stdout) as { status?: unknown; pid?: unknown }
|
||||||
|
const status = typeof parsed.status === 'string' ? parsed.status : null
|
||||||
|
const pid = typeof parsed.pid === 'number' && Number.isInteger(parsed.pid) && parsed.pid > 0
|
||||||
|
? parsed.pid
|
||||||
|
: null
|
||||||
|
return { status, pid }
|
||||||
|
} catch {
|
||||||
|
return { status: null, pid: null }
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
async function waitFor(
|
||||||
|
check: () => Promise<boolean> | boolean,
|
||||||
|
timeoutMs: number,
|
||||||
|
message: string
|
||||||
|
): Promise<void> {
|
||||||
|
const deadline = Date.now() + timeoutMs
|
||||||
|
|
||||||
|
while (Date.now() < deadline) {
|
||||||
|
if (await check()) {
|
||||||
|
return
|
||||||
|
}
|
||||||
|
await sleep(pollIntervalMs)
|
||||||
|
}
|
||||||
|
|
||||||
|
throw new Error(message)
|
||||||
|
}
|
||||||
|
|
||||||
|
console.log('=== Daemon Stop (supervisor regression) ===\n')
|
||||||
|
|
||||||
|
const port = 10000 + Math.floor(Math.random() * 50000)
|
||||||
|
const paseoHome = await mkdtemp(join(tmpdir(), 'paseo-stop-supervisor-'))
|
||||||
|
const cliRoot = join(import.meta.dirname, '..')
|
||||||
|
|
||||||
|
let supervisorProcess: ChildProcess | null = null
|
||||||
|
let recentSupervisorLogs = ''
|
||||||
|
|
||||||
|
try {
|
||||||
|
console.log('Test 1: start daemon-runner in dev mode with isolated PASEO_HOME')
|
||||||
|
|
||||||
|
supervisorProcess = spawn('npx', ['tsx', '../server/scripts/daemon-runner.ts', '--dev'], {
|
||||||
|
cwd: cliRoot,
|
||||||
|
env: {
|
||||||
|
...process.env,
|
||||||
|
PASEO_HOME: paseoHome,
|
||||||
|
PASEO_LISTEN: `127.0.0.1:${port}`,
|
||||||
|
PASEO_RELAY_ENABLED: 'false',
|
||||||
|
CI: 'true',
|
||||||
|
},
|
||||||
|
stdio: ['ignore', 'pipe', 'pipe'],
|
||||||
|
})
|
||||||
|
|
||||||
|
supervisorProcess.stdout?.on('data', (chunk) => {
|
||||||
|
recentSupervisorLogs = (recentSupervisorLogs + chunk.toString()).slice(-8000)
|
||||||
|
})
|
||||||
|
supervisorProcess.stderr?.on('data', (chunk) => {
|
||||||
|
recentSupervisorLogs = (recentSupervisorLogs + chunk.toString()).slice(-8000)
|
||||||
|
})
|
||||||
|
|
||||||
|
await waitFor(
|
||||||
|
async () => {
|
||||||
|
const status = await readDaemonStatus(paseoHome)
|
||||||
|
return status.status === 'running' && status.pid !== null && isProcessRunning(status.pid)
|
||||||
|
},
|
||||||
|
120000,
|
||||||
|
'daemon did not become running in time'
|
||||||
|
)
|
||||||
|
|
||||||
|
const statusBeforeStop = await readDaemonStatus(paseoHome)
|
||||||
|
const daemonPid = statusBeforeStop.pid
|
||||||
|
assert.strictEqual(statusBeforeStop.status, 'running', 'daemon should be running before stop')
|
||||||
|
assert(daemonPid !== null, 'daemon pid should exist once daemon starts')
|
||||||
|
assert(isProcessRunning(daemonPid), 'daemon process should be running')
|
||||||
|
const pidLockBeforeStop = await readPidLockState(paseoHome)
|
||||||
|
assert.strictEqual(pidLockBeforeStop.pid, daemonPid, 'pid lock should match status pid')
|
||||||
|
const command = readProcessCommand(daemonPid)
|
||||||
|
assert(command !== null, 'pid lock pid should resolve to a running process command')
|
||||||
|
assert(
|
||||||
|
command.includes('daemon-runner.ts') || command.includes('daemon-runner.js'),
|
||||||
|
`pid lock pid should be daemon-runner process, got: ${command}`
|
||||||
|
)
|
||||||
|
console.log(`✓ dev daemon started with daemon pid ${daemonPid}\n`)
|
||||||
|
|
||||||
|
console.log('Test 2: `paseo daemon stop` should stop without respawn')
|
||||||
|
const stopResult =
|
||||||
|
await $`PASEO_HOME=${paseoHome} npx paseo daemon stop --home ${paseoHome} --json`.nothrow()
|
||||||
|
assert.strictEqual(stopResult.exitCode, 0, `stop should succeed: ${stopResult.stderr}`)
|
||||||
|
const stopJson = JSON.parse(stopResult.stdout) as { action?: unknown }
|
||||||
|
assert.strictEqual(stopJson.action, 'stopped', 'stop should report stopped action')
|
||||||
|
|
||||||
|
await waitFor(async () => {
|
||||||
|
const status = await readDaemonStatus(paseoHome)
|
||||||
|
return status.status === 'stopped'
|
||||||
|
}, 15000, 'daemon status did not transition to stopped after stop')
|
||||||
|
|
||||||
|
if (supervisorProcess.pid) {
|
||||||
|
await waitFor(
|
||||||
|
() => !isProcessRunning(supervisorProcess!.pid ?? -1),
|
||||||
|
15000,
|
||||||
|
'daemon-runner supervisor remained running after stop'
|
||||||
|
)
|
||||||
|
}
|
||||||
|
|
||||||
|
await sleep(1000)
|
||||||
|
|
||||||
|
const pidAfterStop = await readPidLockState(paseoHome)
|
||||||
|
const respawned = pidAfterStop.pid !== null && isProcessRunning(pidAfterStop.pid)
|
||||||
|
assert.strictEqual(
|
||||||
|
respawned,
|
||||||
|
false,
|
||||||
|
`daemon respawned after stop (pid: ${pidAfterStop.pid ?? 'unknown'})`
|
||||||
|
)
|
||||||
|
|
||||||
|
const statusAfterStop = await readDaemonStatus(paseoHome)
|
||||||
|
assert.strictEqual(statusAfterStop.status, 'stopped', 'daemon should remain stopped after stop command')
|
||||||
|
console.log('✓ stop leaves supervised daemon stopped (no respawn)\n')
|
||||||
|
} finally {
|
||||||
|
if (supervisorProcess?.pid && isProcessRunning(supervisorProcess.pid)) {
|
||||||
|
supervisorProcess.kill('SIGTERM')
|
||||||
|
await waitFor(() => !isProcessRunning(supervisorProcess!.pid ?? -1), 5000, 'supervisor cleanup timed out').catch(
|
||||||
|
() => {
|
||||||
|
supervisorProcess?.kill('SIGKILL')
|
||||||
|
}
|
||||||
|
)
|
||||||
|
}
|
||||||
|
|
||||||
|
await $`PASEO_HOME=${paseoHome} npx paseo daemon stop --home ${paseoHome} --force`.nothrow()
|
||||||
|
await rm(paseoHome, { recursive: true, force: true })
|
||||||
|
}
|
||||||
|
|
||||||
|
if (recentSupervisorLogs.trim().length === 0) {
|
||||||
|
// Keep output stable while still surfacing that logs were captured when needed.
|
||||||
|
console.log('(no supervisor logs captured)')
|
||||||
|
}
|
||||||
|
|
||||||
|
console.log('=== Supervisor stop regression test passed ===')
|
||||||
@@ -34,6 +34,17 @@ function getPidFilePath(paseoHome: string): string {
|
|||||||
return join(paseoHome, "paseo.pid");
|
return join(paseoHome, "paseo.pid");
|
||||||
}
|
}
|
||||||
|
|
||||||
|
function resolveLockOwnerPid(): number {
|
||||||
|
if (typeof process.send === "function") {
|
||||||
|
const ppid = process.ppid;
|
||||||
|
if (Number.isInteger(ppid) && ppid > 1) {
|
||||||
|
return ppid;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
return process.pid;
|
||||||
|
}
|
||||||
|
|
||||||
export async function acquirePidLock(
|
export async function acquirePidLock(
|
||||||
paseoHome: string,
|
paseoHome: string,
|
||||||
sockPath: string
|
sockPath: string
|
||||||
@@ -55,8 +66,13 @@ export async function acquirePidLock(
|
|||||||
}
|
}
|
||||||
|
|
||||||
// Check if existing lock is stale
|
// Check if existing lock is stale
|
||||||
|
const lockOwnerPid = resolveLockOwnerPid();
|
||||||
if (existingLock) {
|
if (existingLock) {
|
||||||
if (isPidRunning(existingLock.pid)) {
|
if (isPidRunning(existingLock.pid)) {
|
||||||
|
if (existingLock.pid === lockOwnerPid) {
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
throw new PidLockError(
|
throw new PidLockError(
|
||||||
`Another Paseo daemon is already running (PID ${existingLock.pid}, started ${existingLock.startedAt})`,
|
`Another Paseo daemon is already running (PID ${existingLock.pid}, started ${existingLock.startedAt})`,
|
||||||
existingLock
|
existingLock
|
||||||
@@ -68,7 +84,7 @@ export async function acquirePidLock(
|
|||||||
|
|
||||||
// Create new lock with exclusive flag
|
// Create new lock with exclusive flag
|
||||||
const lockInfo: PidLockInfo = {
|
const lockInfo: PidLockInfo = {
|
||||||
pid: process.pid,
|
pid: lockOwnerPid,
|
||||||
startedAt: new Date().toISOString(),
|
startedAt: new Date().toISOString(),
|
||||||
hostname: hostname(),
|
hostname: hostname(),
|
||||||
uid: process.getuid?.() ?? 0,
|
uid: process.getuid?.() ?? 0,
|
||||||
@@ -103,11 +119,12 @@ export async function acquirePidLock(
|
|||||||
|
|
||||||
export async function releasePidLock(paseoHome: string): Promise<void> {
|
export async function releasePidLock(paseoHome: string): Promise<void> {
|
||||||
const pidPath = getPidFilePath(paseoHome);
|
const pidPath = getPidFilePath(paseoHome);
|
||||||
|
const lockOwnerPid = resolveLockOwnerPid();
|
||||||
try {
|
try {
|
||||||
// Only remove if it's our lock
|
// Only remove if it's our lock
|
||||||
const content = await readFile(pidPath, "utf-8");
|
const content = await readFile(pidPath, "utf-8");
|
||||||
const lock = JSON.parse(content) as PidLockInfo;
|
const lock = JSON.parse(content) as PidLockInfo;
|
||||||
if (lock.pid === process.pid) {
|
if (lock.pid === lockOwnerPid) {
|
||||||
await unlink(pidPath);
|
await unlink(pidPath);
|
||||||
}
|
}
|
||||||
} catch {
|
} catch {
|
||||||
|
|||||||
Reference in New Issue
Block a user