Connection Pool Size Limits in Meteor: Unresolved Exhaustion Behavior

Answer

1. Maximum reliable pool size: Meteor does not enforce an upper limit; the practical limit is set by the MongoDB server’s max connections and the Node.js driver. Values in the range 10‑200 have been used stably in Meteor 1.12+, but exceeding the server’s capacity leads to connection refusals and increased latency.
2. Setting MONGO_CONNECTION_POOL_SIZE above the default does not cause Meteor to crash; instability only appears if the MongoDB server cannot accommodate the extra connections, resulting in timeout errors or refused connections.
3. Meteor itself provides no automatic back‑pressure or retry beyond the driver’s internal queue. Operations wait for a free connection up to waitQueueTimeoutMs; after that a MongoError is thrown and must be handled by application code.

Confirmed Facts

• Meteor relies on the node‑mongodb-native driver; the pool size is controlled by the maxPoolSize option in the MONGO_URL (default 5 connections unless overridden). Meteor’s own startup code adds a baseline of 10 connections when the variable is not set.
• When the pool is exhausted, incoming MongoDB operations are queued; they block until a connection becomes free or until waitQueueTimeoutMs expires, after which a MongoError with message 'Timed out while checking out connection from connection pool' is thrown.
• Exhaustion does not crash the app; it manifests as increased latency, queued requests, and eventual timeouts.

Likely Explanation

The observed “connection pool exhausted” errors occur when the rate of MongoDB requests exceeds the rate at which connections are returned to the pool. Because Meteor does not apply back‑pressure, the queue grows until waitQueueTimeoutMs is hit, producing the error seen by clients.

Steps to Verify

1. Check the effective MONGO_URL: console.log(process.env.MONGO_URL) or inspect MongoInternals.defaultRemoteCollectionDriver().mongo.clientOptions at startup.
2. Run db.serverStatus().connPoolStats on the MongoDB instance under load to see active, available, and waiting connections.
3. If you see frequent timeout errors, note the current waitQueueTimeoutMs value (default is 0 meaning no timeout). Adjusting it may change whether you get queued delays versus immediate errors.

Missing Diagnostic Detail

To decide whether to increase the pool size or lower waitQueueTimeoutMs, confirm the observed timeout pattern: are you seeing queued delays that later succeed, or immediate timeout errors? This detail determines whether raising the pool size or tuning the timeout is the appropriate next step.