Skip to content

Investigate catchall alerts and pgbouncer saturation

@marcogreg called this out in the incident and I wanted to track that it exists.

Over the past week, the EOC has been notified multiple times about catchall sidekiq_queueing SLI, which has autoresolved in every case.

The sidekiq PG bouncer pool was decreased on September 24th: gitlab-com/gl-infra/production#20505 (closed)

Since then, we're seeing an increase in pgbouncer saturation and queueing apdex.

source

image__1_

image

We should probably either increase the sidekiq pgbouncer pool back or somehow tweak that queueing SLI to stop paging.

This ticket was created from INC-4634 and was automatically exported by incident.io 🔥

Edited by Stephanie Jackson
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information