RELEASED: Public: May 28, 2025
New features
- HA controller maximum surge during rolling restart (BEE-58041)
-
Previously, an HA controller’s
Deployment
used the Kubernetes defaultmaxSurge
value of 25%, which was inappropriate since, in fact, only one new replica can start at a time; others would wait to acquire a startup lock. Now it is set to one.
maxUnavailable
is also set to zero by default, so running replicas are not removed until replacements are ready. In clusters with extremely constrained resources, this setting may prevent a rolling restart from occurring if the configured replica count is the maximum that can be scheduled. In that case, you may need to override custom YAML to tolerate a temporary scale down:
apiVersion: apps/v1 kind: Deployment spec: strategy: rollingUpdate: maxUnavailable: 1