We know of about at least 3 clusters where the liveness probes failed due to the VPA recommender overwhelming the Prometheus web server (max TCP connections reached) and leading to continuous restarts of Prometheus pods.

• OCPBUGS-18971
• OCPBUGS-15337
• OCPBUGS-4186 (suspicion, to be confirmed)

We need to describe exhaustively the issue and how we can mitigate it. OCP already has a post-mortem template that we can use.

DoD

• Post-mortem document reviewed by the team.