(not always bizarre, sometimes just backfiring daily design)
HorizontalPodAutoscaler
takes into accountTerminating
pods when calculating the metrics. The consequence is if a pod has a bug on its termination procedure, terminating pods as part of a scale down triggers a scale up.- It once did the same for
Pending
pods.
- It once did the same for
- Can't omit
clusterIP
if you ever set it (even if to "") https://gitlab.com/gitlab-org/charts/gitlab/issues/1710#note_262760589 - How do you force HTTPS redirection on Ambassador: you use a nginx sidecar
- How do you force static HTTP responses (401, for example) on Ambassador: you use a nginx sidecar