Skip to content

component_saturation_slo_out_of_bounds:pgbouncer_single_core

The pgbouncer_single_core metric measures the total CPU usage by PGBouncer process. PGBouncer is a single threaded application, so if the process fully consumes CPU, a resource may become saturated. This may cause slowness or timeouts on the DB requests. As a result additional pgbouncer nodes may need to be provisioned.

This alert can cause a high Severity incident and needs to be investigated.

  • The alert temlpate expression measures the total CPU usage by PGBouncer process. This template is used to auto-generate alert templates for different environments and Service type (Patroni or Pgbouncer)
  • Generated alert measures the CPU usage rate of pgbouncer processes, then sums this usage across CPUs and modes and clamps the result between 0 and 1 (0-100% CPU usage). Examples of generated alerts for Patroni and for Pgbouncer
  • The alert has soft and hard SLO defined
  • This alert should be rare, but if it’s triggered, needs to be investigated immediately
  • This alert might create S2 incidents
  • There might be many gitlab.com users affected as the requrest to backend DB may be delayed
  • Review Incident Severity Handbook page to identify the required Severity Level

Under normal conditions the saturation is expected to be as small as possible. The closer the saturation to 1, the higher the load and new additional nodes are required to be provisioned.

  • There are no external or internal dependencies for this alert

For escalation contact the following channels:

Alternative slack channels: