Scales Kubernetes worker nodes within autoscaling groups.

Automatically adjust resources for your workloads

Changes by Kind

API Change

• Added MemoryAggregationIntervalSeconds for the VPA object (#9317, @omerap12)
• EvictAfterOOMThreshold is now per vpa (#9071, @omerap12)
• The VPA support InPlace update mode. VPA will not evict any pods and only try to resize pods inplace. (#8888, @omerap12)
• Users can configure a startupBoost policy in the VPA spec. (#9141, @kamarabbas99)
• VPA now stores a observedGeneration in both status and condition fields. (#9524, @adrianmoisey)

Bug or Regression

• Fix bug where VPA admission-controller could write a webhook with an empty cert for a brief period. (#9599, @adrianmoisey)
• Fixed OOM detection for pods with restartPolicy: Never (#7955, @tkukushkin)
• Fixed a bug where the presence of a LimitRange object with CPU constraints in the cluster prevented VPA from recreating Pods managed by the targeted controllers. The updated code section, which proportionally reduces resource requests and limits, could result in the loss of sub-millicore values. (#8946, @iamzili)
• Fixed a bug where users could use alpha (off-by-default) features by updating an existing VPA resource, bypassing the feature gate check. (#9557, @adrianmoisey)

Other (Cleanup or Flake)

• Bump Kubernetes dependencies to 1.36.0 (#9547, @adrianmoisey)
• Reduce VPA memory usage by stripping away unnecessary managed fields from informer cache (#9634, @adrianmoisey)
• Update flag in-place-skip-disruption-budget is now beta (#9655, @omerap12)
• Improved VPA validation error messages now pinpoint exactly which fields are invalid, making it faster to diagnose and resolve configuration issues in your VerticalPodAutoscaler specs. (#9591, @adrianmoisey)

Automatically adjust resources for your workloads

Scales Kubernetes worker nodes within autoscaling groups.

Scales Kubernetes worker nodes within autoscaling groups.

Automatically adjust resources for your workloads

Scales Kubernetes worker nodes within autoscaling groups.

What's Changed

• [cluster-autoscaler-release-1.32] Update k8s deps to v1.32.12 by @jackfrancis in #9231
• [cluster-autoscaler-release-1.32] GOARCH as --build-arg for docker buildx by @jackfrancis in #9233

Full Changelog: cluster-autoscaler-1.32.6...cluster-autoscaler-1.32.7

Images

• registry.k8s.io/autoscaling/cluster-autoscaler:v1.32.7
• registry.k8s.io/autoscaling/cluster-autoscaler-arm64:v1.32.7
• registry.k8s.io/autoscaling/cluster-autoscaler-amd64:v1.32.7
• registry.k8s.io/autoscaling/cluster-autoscaler-s390x:v1.32.7

Changes by Kind

API Change

• Correctly mark the VPA UpdateMode "Auto" as deprecated, that was deprecated in VPA 1.5 (#9073, @adrianmoisey)

Feature

• Adds --in-place-skip-disruption-budget flag that skips disruption budget checks for in-place pod updates when all containers have NotRequired resize policy (#8987, @omerap12)
• Promote InPlaceOrRecreate feature to GA, defaulted to enabled (#9082, @adrianmoisey)

Bug or Regression

• Avoid returning error when annotation patch for InPlaceOrResize fails, that would cause Pods to get unnecessarily evicted after resize (#9115, @adrianmoisey)
• Optimize memory usage in VPA components by scoping informers to configured namespaces. (#9122, @adrianmoisey)
• Vertical-pod-autoscaler: Fix the version of the k8s.io/client-go module. (#8579, @ialidzhikov)
• Vpa-updater and vpa-admission-controller no longer excessively log fail to get pod controller: (...) last error node is not a valid owner (#8611, @adrianmoisey)
• VPA: Fix recommender race conditions for vpa Conditions and Recommendations (#8967, @jkyros)

Other (Cleanup or Flake)

• Update Kubernetes libraries to v1.35.0 (#8955, @adrianmoisey)