n8n - 💡(How to fix) Fix job.finished() leaks EventEmitter listeners and setInterval in queue mode when removeOnComplete/removeOnFail is true [1 comments, 2 participants]

Bug Description

In queue mode, webhook/main pods permanently leak Bull job.finished() resources when a global:completed or global:failed event is missed (e.g. during a Redis reconnect). Over time this causes the process to OOM-crash.

Root Cause

WorkflowRunner.enqueueExecution() calls job.finished() to wait for the worker to complete a Bull job. Internally, Bull's job.finished() does three things:

1. Adds a global:completed listener to the queue EventEmitter
2. Adds a global:failed listener to the queue EventEmitter
3. Starts a watchdog setInterval every 5 seconds that polls scripts.isFinished()

These are cleaned up by removeListeners() only when a completion event fires.

scaling.service.ts sets removeOnComplete: true and removeOnFail: true. This means completed/failed jobs are immediately removed from Redis. If the global:completed/global:failed event is missed (Redis reconnect, timing issue), the watchdog calls scripts.isFinished() — but the job no longer exists in Redis, so it returns -1 (not found). The watchdog interprets this as "still running" and keeps polling forever. removeListeners() is never called.

Each affected job permanently retains:

• 2 Bull EventEmitter listeners on the queue object
• 1 setInterval handle (5s polling, never cleared)
• The activeExecutions[executionId] entry (closure holds full executionData, request body, and HTTP response socket via httpResponse)

Impact

• Webhook pods OOM-crash on a daily basis under production load
• Each leaked entry holds the full execution payload + open HTTP socket
• Confirmed present in all 1.x versions and 2.x up to 2.20.7

To Reproduce

1. Run n8n in queue mode with separate webhook and worker pods
2. Trigger a brief Redis reconnection while jobs are in flight (or under sustained high load where events can be occasionally missed)
3. Monitor Object.keys(activeExecutions).length on the webhook pod — it grows monotonically and never shrinks

Expected behavior

job.finished() should resolve (or reject) for every job regardless of whether removeOnComplete/removeOnFail is set, and all associated EventEmitter listeners and timers should always be cleaned up after a job completes or fails.

Debug Info

core

• n8nVersion: 1.123.26
• platform: docker (self-hosted)
• nodeJsVersion: 24.13.1
• nodeEnv: production
• database: postgres
• executionMode: scaling (multi-main)
• concurrency: -1
• license: enterprise (sandbox)
• consumerId: a891fcab-3387-4047-ba6a-a954540e45f1

storage

• success: none
• error: none
• progress: false
• manual: true
• binaryMode: memory

pruning

• enabled: true
• maxAge: 336 hours
• maxCount: 2000000 executions

client

• userAgent: mozilla/5.0 (macintosh; intel mac os x 10_15_7) applewebkit/537.36 (khtml, like gecko) chrome/148.0.0.0 safari/537.36
• isTouchDevice: false

security

• secureCookie: false

Generated at: 2026-05-13T10:50:34.320Z

Operating System

node:24.13.1-alpine

n8n Version

1.123.26

Node.js Version

24.13.1

Database

PostgreSQL

Execution mode

queue

Hosting

self hosted