Distributed lock mongodb

Distributed Lock over MongoDB

Shape

A DistributedLockService exposes tryLock(key, ttl) / release(key, owner).

Implementation uses atomic findAndModify on a single locks collection:

• Document: { _id: key, owner: "<node>:<thread>", expiresAt: Date }
• Acquire: findAndModify(query = { _id: key, $or: [{expiresAt: {$lt: now}}, {_id: {$exists: false}}] }, update = { owner, expiresAt: now + ttl }, upsert: true)
• Release: deleteOne({ _id: key, owner: self }) (never blind delete)
• Retry: loop with 100 ms sleep up to configured timeout before giving up

Steps

1. Create collection locks with TTL index on expiresAt as a safety net (not authoritative — acquire query checks expiry).
2. Implement tryLock using Mongo atomic upsert; return false when a different non-expired owner holds the key.
3. Always stamp owner = "<hostname>:<threadName>" so release is idempotent and crash-safe.
4. Wrap critical section in try { if (tryLock) ... } finally { release }.
5. For retry-until-timeout semantics, add a thin loop around tryLock rather than blocking inside the lock method.

Counter / Caveats

• NTP drift across nodes skews TTL comparison — keep TTLs generous (seconds, not ms).
• Do not use a TTL index as the only expiry mechanism; the background deleter runs every 60 s.
• Prefer Redis/Redisson if you need fair queuing or pub-sub release notifications.