Hi, we are experiencing a problem with the postgres volume filling up very quickly, not related to data but related to logs. The root cause seems that Dagster is connecting many times per second and postgres logging all the (re)connections. Example log:

2022-06-13 23:59:59.727 UTC,"dagster","dagster",2251490,"10.42.80.16:59264",62a7cf7f.225ae2,3,"idle",2022-06-13 23:59:59 UTC,,0,LOG,00000,"disconnection: session time: 0:00:00.011 user=dagster database=dagster host=10.42.80.16 port=59264",,,,,,,,,""

It is possible to disable this logging in Postgres, but not recommended. We also tried configuring this through dagster but we didn’t find any options. Do you have any best practices to prevent this from happening?

Hi Pieter. What version of postgres and what version of Dagster are you using?

Dagster 0.14.20, postgres 14

Hi Pieter. It looks like we are using connection pooling for some class of queries, but not in all places (e.g. we’re creating new connections when you view a run, for example)…. I can try to track down a solution, but I’ll create an issue for now since it might not be a quick fix.

Created issue at: https://github.com/dagster-io/dagster/issues/8466

Thanks a lot!

Hi @prha any news on this? We are facing similar problems

No new updates… but I’ll try to spend some more hours on this issue this week