Unable to acquire connection #145
Description
I seem to be hitting a regression on v1.17.2, possibly related to #143.
On startup, the relay will process requests anywhere from a minute to an hour, after which point the relay stalls and spams the log with this error: Unable to acquire connection. Please increase DB_MAX_POOL_SIZE, DB_ACQUIRE_CONNECTION_TIMEOUT and tune postgresql.conf to make use of server's resources.
Over the course of today I've tried to tune pool size, ultimately to no effect. No matter what I do, after about an hour (or less) all of the workers stall out and the relay stops responding to clients altogether.
When I look at the pg_stat_activity table, all of the pool sessions are still open. I'm using watch "uptime && psql nostream -c 'select datid,datname,pid,usename,age(now(),query_start),state_change,wait_event_type,wait_event,state,backend_xid,backend_xmin,backend_type from pg_stat_activity;'"
Rolling back to v1.16.0 appears to resolve the issue.
- OS: Ubuntu 22.04
- Platform: docker 20.10.22
- ts-relay Version: 1.17.2
- Postgresql 13.9-1, and 14.6-1
Happy to attach more debug logs if it helps. I can just roll forward again for a bit.