Distributed lock implementation

export interface DistributedLock {
  acquire(): Promise<boolean>;
  release(): Promise<void>;
  renew(): Promise<boolean>;
  isAcquired(): boolean;
}

import Redis from 'ioredis';

export class RedisDistributedLock implements DistributedLock {
  private readonly lockKey: string;
  private readonly lockValue: string;
  private readonly ttlMs: number;
  private readonly renewalInterval: number;
  private acquired = false;
  private renewalTimer: NodeJS.Timeout | null = null;

  constructor(
    private readonly redis: Redis,
    key: string,
    ttlMs = 10000,
    renewalInterval = 3000
  ) {
    this.lockKey = `lock:${key}`;
    this.lockValue = `${process.pid}:${crypto.randomUUID()}`;
    this.ttlMs = ttlMs;
    this.renewalInterval = renewalInterval;
  }

  async acquire(): Promise<boolean> {
    const lua = `
      if redis.call('SET', KEYS[1], ARGV[1], 'NX', 'PX', ARGV[2]) == 1 then
        return 1
      else
        return 0
      end
    `;

    const result = await this.redis.eval(lua, 1, this.lockKey, this.lockValue, this.ttlMs);
    this.acquired = result === 1;

    if (this.acq

uired) { this.startRenewal(); }

return this.acquired;

}

async release(): Promise<void> { if (!this.acquired) return;

const lua = `
  if redis.call('GET', KEYS[1]) == ARGV[1] then
    return redis.call('DEL', KEYS[1])
  else
    return 0
  end
`;

await this.redis.eval(lua, 1, this.lockKey, this.lockValue);
this.acquired = false;
this.stopRenewal();

}

async renew(): Promise<boolean> { const lua = if redis.call('GET', KEYS[1]) == ARGV[1] then return redis.call('PEXPIRE', KEYS[1], ARGV[2]) else return 0 end ;

const result = await this.redis.eval(lua, 1, this.lockKey, this.lockValue, this.ttlMs);
return result === 1;

}

isAcquired(): boolean { return this.acquired; }

private startRenewal(): void { this.renewalTimer = setInterval(async () => { const renewed = await this.renew(); if (!renewed) { this.acquired = false; this.stopRenewal(); } }, this.renewalInterval); }

private stopRenewal(): void { if (this.renewalTimer) { clearInterval(this.renewalTimer); this.renewalTimer = null; } } }

**3. Contention handling wrapper**
```typescript
export async function withDistributedLock<T>(
  redis: Redis,
  key: string,
  task: () => Promise<T>,
  options = { maxRetries: 3, baseDelayMs: 100, ttlMs: 10000 }
): Promise<T> {
  const lock = new RedisDistributedLock(redis, key, options.ttlMs);

  for (let attempt = 0; attempt <= options.maxRetries; attempt++) {
    const acquired = await lock.acquire();
    if (acquired) {
      try {
        return await task();
      } finally {
        await lock.release();
      }
    }

    if (attempt === options.maxRetries) {
      throw new Error(`Failed to acquire lock after ${options.maxRetries} retries`);
    }

    const delay = options.baseDelayMs * Math.pow(2, attempt) + Math.random() * 100;
    await new Promise(resolve => setTimeout(resolve, delay));
  }

  throw new Error('Unreachable');
}

Rationale

• Lua atomicity prevents the check-then-act race condition that breaks naive implementations
• Background renewal decouples task duration from lock TTL, eliminating premature expiration
• Exponential backoff with jitter prevents thundering herd scenarios during high contention
• Ownership verification ensures processes only release locks they hold, critical in GC pause scenarios where a process might hold a lock past its TTL

Pitfall Guide

1. Using SETNX without TTL Locks never expire when processes crash or hang. The system deadlocks until manual intervention. Always pair acquisition with PX or EX to enforce lease semantics.

2. Releasing locks without ownership verification A process that acquires a lock, experiences a GC pause past the TTL, and then attempts to release will delete a lock now held by another process. Lua scripts must verify the lock value matches the owner UUID before deletion.

3. Ignoring clock skew across nodes Redis TTL relies on server time. In multi-node deployments, clock drift causes premature expiration or extended holds. Use lease renewal and prefer consensus-based systems (etcd, ZooKeeper) when strict temporal guarantees are required.

4. No lease renewal for long-running tasks Fixed TTLs assume predictable execution time. Background workers processing large payloads or waiting on external APIs will exceed TTLs. Implement automatic renewal at half the TTL interval.

5. Single-node lock services in production A single Redis instance becomes a single point of failure. Network partitions cause split-brain scenarios where multiple nodes believe they hold the same lock. Deploy Redis Sentinel or Cluster, or use quorum-based acquisition (Redlock) for critical paths.

6. Blocking retries without jitter Synchronous retry loops with fixed delays cause thundering herd effects. All contending processes wake simultaneously, overwhelming the lock service. Add randomized jitter to backoff calculations.

7. Treating locks as transaction boundaries Distributed locks coordinate access, not guarantee consistency. They do not replace idempotency keys, optimistic concurrency control, or compensating transactions. Locks should protect critical sections, not entire business workflows.

Best Practices from Production

• Monitor lock acquisition latency and contention rates; alert when >5% of acquisitions require retries
• Use circuit breakers for the lock service to prevent cascading failures during Redis outages
• Set TTL to 3–5x the expected critical section duration; renewal handles variance
• Never nest distributed locks across different keys without a strict ordering protocol to prevent deadlocks
• Log lock lifecycle events (acquire, renew, release, timeout) with trace IDs for distributed tracing

Production Bundle

Action Checklist

• Define lock TTL based on worst-case critical section duration plus 200% buffer
• Implement Lua-based acquisition with SET NX PX and unique owner UUID
• Add background lease renewal at 30–50% of TTL interval
• Verify ownership before release using atomic Lua scripts
• Configure exponential backoff with jitter for contention retries
• Deploy lock service with high availability (Sentinel/Cluster or etcd cluster)
• Instrument acquisition latency, renewal success rate, and contention metrics
• Add circuit breaker fallback to degrade gracefully during lock service outages

Decision Matrix

Scenario	Recommended Approach	Why	Cost Impact
Non-critical worker deduplication	Single-Node Redis + TTL	Low latency, simple deployment, acceptable failure risk	Low infrastructure cost
Financial transaction coordination	etcd Lease or Redis Redlock	Partition tolerance, clock skew resilience, strong consistency	Moderate infrastructure cost
Serverless function coordination	Redis Cluster + Quorum Acquisition	Stateless functions require external lease management with high availability	Pay-per-use Redis cluster cost
Database-backed critical paths	Advisory Locks + Application-level retry	Leverages existing DB ACID guarantees, avoids external dependency	Zero additional infrastructure

Configuration Template

// lock.config.ts
import Redis from 'ioredis';

export const lockRedis = new Redis({
  host: process.env.REDIS_HOST || 'localhost',
  port: parseInt(process.env.REDIS_PORT || '6379'),
  password: process.env.REDIS_PASSWORD,
  maxRetriesPerRequest: 3,
  retryStrategy: (times) => Math.min(times * 50, 2000),
  enableReadyCheck: true,
  reconnectOnError: (err) => {
    const targetError = 'READONLY';
    if (err.message.includes(targetError)) {
      return true;
    }
    return false;
  }
});

export const lockDefaults = {
  ttlMs: 15000,
  renewalInterval: 5000,
  maxRetries: 4,
  baseDelayMs: 150,
  jitterRange: 200
};

export function createLock(key: string) {
  const { RedisDistributedLock } = require('./distributed-lock');
  return new RedisDistributedLock(
    lockRedis,
    key,
    lockDefaults.ttlMs,
    lockDefaults.renewalInterval
  );
}

Quick Start Guide

1. Install dependencies: npm install ioredis uuid
2. Create lock instance: const lock = createLock('order-processing:12345');
3. Acquire and execute:

   const acquired = await lock.acquire();
   if (acquired) {
     try {
       // critical section
     } finally {
       await lock.release();
     }
   }
   

4. Wrap with retry: Use withDistributedLock(redis, key, task) for automatic backoff and cleanup
5. Verify in monitoring: Check Redis keyspace for lock:* patterns and confirm TTL expiration behavior under load