n8n errors / Connectivity
Connection refused on SSH-tunnelled Postgres (127.0.0.1:34827)
Cause
n8n v1.78.1 introduced connection pooling with a hard maxConnections default that causes SSH-tunnelled Postgres connections to time out and refuse in looping workflows.
Fix
Upgrade to n8n v1.99.0 which contains the SSH-tunnel connection-pool fix. On earlier versions, raise the SSH tunnel idle timeout and reduce parallelism in loops that hit Postgres so the pool is not exhausted.
Part of the Automation Error Index — n8n errors: browse all.
Hitting an automation error we have not covered?
AutomateLab builds and debugs n8n, Make, and agent workflows as fixed-scope projects. If something is broken in production, tell us what you are seeing.
Describe the problem