mmk unlock

Force-release a stuck concurrency lock left behind by a crashed migration run.

mmk unlock [options]

Why it exists

Every write command acquires an atomic MongoDB lock (in the _mmk_locks collection) so two deploys can't run migrations at once. The lock auto-expires after lockTTLSeconds and renews on a heartbeat while a migration runs. But if a process is killed mid-run, a lock can linger until its TTL elapses — mmk unlock clears it immediately.

Usage

mmk unlock          # show the current holder, then prompt for confirmation
mmk unlock --yes    # force-release without prompting
mmk unlock --json   # machine-readable { released, holder }

It first shows who holds the lock:

Lock held by:
  pid:   48213
  host:  deploy-runner-7
  user:  ci
  since: 2026-06-05 12:01:33

Release this lock? [y/N]

Options

Option	Description
--yes	Skip the confirmation prompt and release immediately.
--json	Emit { released, holder } as JSON.

Plus the global flags.

Only when you're sure

Force-releasing a lock that another process still holds can let two runs proceed concurrently. Only unlock when you've confirmed the holding process is actually dead.

Programmatic API

The same capability is available on MigratorKit:

const info = await migrator.lockInfo();   // → LockInfo | null
if (info) {
  await migrator.forceUnlock();
}