Node.js Memory Leak That Required Daily Restarts

A closures-based event listener accumulation caused heap growth of 40 MB/hour, requiring nightly automated restarts to prevent OOM crashes.

Overview

A Node.js API crashed every night around 3 AM for months because every request added a listener to a shared Redis client and never removed it.

What happened

The nightly restart took only 30 seconds and hit low traffic, so engineers accepted it as normal. Scaling out made the pattern impossible to ignore.

Why it broke

Each request registered a close listener on the shared Redis client to handle request teardown. The listener was never removed, and its closure kept the whole request object alive.

Restart time — 30 seconds
Instances before and after scaling — 2, then 8
Snapshot interval that exposed the leak — 4 hours
Listeners after 10,000 requests — 10,000

The fix

One line of cleanup, plus a guard that makes the next leak loud instead of silent.

Listener lifetime — Forever — Removed when the response ends
Leak visibility — A nightly restart nobody questioned — Warning when listeners pass the threshold

> [!NOTE] A crash that a health check "fixes" every night is still a crash. Scaling out turned a quiet 3 AM restart into daytime outages.

Key concepts

Heap snapshot: A point-in-time dump of the JavaScript heap showing all allocated objects, their sizes, and reference chains — used to identify what is growing over time.
Event listener accumulation: A common Node.js leak pattern where listeners are added to emitters without corresponding removeListener calls, causing references to accumulate.
Closure reference leak: A function (closure) that captures outer-scope variables keeps those variables alive in memory even after the outer scope should have been garbage collected.
MaxListeners warning: Node.js emits a warning when more than 10 listeners are added to a single event on an emitter, indicating a potential leak.

When to use it

Discussing Node.js memory management and closure gotchas
Explaining how to diagnose memory leaks with heap snapshots
Justifying operational observability requirements (heap metrics, process memory)

Interview questions

A Node.js service is growing in memory usage by 50 MB per hour. How do you diagnose and fix it?