Postgres DB mysteriously has dropped important scheduler indexes #29120
-
I use airflow and it's related postgres DB through k9s and I've been seeing a very high CPU utilization % for one of our postgres instances. For the instance with high CPU utilization %, I could identify several airflow queries that look related to normal scheduler tasks that have a higher than expected latency. I use
For example:
has a P50 latency of 139 ms in the problem pod and 2 ms in our regular pod. Upon further investigation, it seems like this is related to dropped indexes in the problematic postgres.
The problematic postgres DB is missing indexes on tables like job, dag and just has indexes on ab_permission, ap_permission_view, ab_user, ab_role and a couple of other ones. Couple of questions:
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
I suspect the indexes being dropped is totally unrelated to any airflow interaction...so main question is really is there a recommended approach for restoring the indexes that should be there? |
Beta Was this translation helpful? Give feedback.
I suspect the indexes being dropped is totally unrelated to any airflow interaction...so main question is really is there a recommended approach for restoring the indexes that should be there?