MongoDB "connection pool ... cleared" / pool errors in CI

The driver cleared its connection pool for a server after a network or handshake failure, then reconnects. Occasional clears during a blip are normal; frequent clears in CI point to an unstable server or too many concurrent connections.

What this error means

Logs show "connection pool for <host> was cleared because of another operation failure" or "connection pool paused", sometimes alongside slow or failing queries.

node

[connection] connection pool for 127.0.0.1:27017 was cleared
because of another operation failure
[connection] connection pool for 127.0.0.1:27017 is now paused

Common causes

A transient server error triggers a pool reset

When one operation fails at the network layer, the driver clears the pool for that server to drop stale sockets, then reconnects.

Too many concurrent connections for a small test mongod

A parallel test suite opens more connections than maxPoolSize or the server allows, causing contention and clears.

How to fix it

Reuse one client and bound the pool

Create a single MongoClient for the whole test run instead of one per test.
Set a reasonable maxPoolSize for the runner size.
Ensure tests do not leak clients that keep opening connections.

db.js

const client = new MongoClient(process.env.MONGODB_URI, {
  maxPoolSize: 10,
  serverSelectionTimeoutMS: 10000,
});

Stabilize the server before load

Wait for the mongo healthcheck and give the container enough memory so it does not stall under the suite.

How to prevent it

Share one client across the suite; do not open a client per test.
Set maxPoolSize appropriate to the runner and server.
Connect only after the server is healthy.