The Wire · Showcase
VERTICAL POD AUTOSCALER 1.7.1 FIXES PROMETHEUS HISTORY REGRESSION
By RepoJournal · Filed · About Kubernetes
VPA's recommender was silently ignoring metric flags when querying Prometheus history, causing recommendations to restart from scratch instead of building on past data.
Vertical Pod Autoscaler 1.7.1 landed with two critical fixes for the recommender [1]. The first addresses a regression where the --history-cpu-metric and --history-memory-metric flags were ignored when using the Prometheus history provider, causing historical queries to skip the configured metric name entirely [1]. The second restores checkpoint loading to prevent recommendations from restarting from scratch [1]. Both regressions would have forced unnecessary recommendation churn in production. The chart version 0.11.0 now defaults to this release [2]. Meanwhile, kube-state-metrics gained two observability improvements: it now logs when a configured custom resource CRD is absent instead of silently generating no metrics [3], and added support for label generation on ephemeral volumes to align with upstream PVC naming conventions [4]. Documentation continues its steady refresh across multiple languages, with Portuguese-Brazilian, French, and Japanese docs migrating to the newer code_sample shortcode [5] [6].
One email a day. Unsubscribe in one click.
Keep up with Kubernetes in about 3 minutes a day: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
One email a day. Unsubscribe in one click. Read a past issue →
Action items
- → Upgrade VPA to 1.7.1 immediately if using Prometheus history provider kubernetes/autoscaler [immediate]
- → Review kube-state-metrics logs for missing CRD warnings after upgrade kubernetes/kube-state-metrics [plan]
- → Verify Manila CSI plugin access rule polling on next update cycle kubernetes/cloud-provider-openstack [monitor]
References
- [1] vertical-pod-autoscaler-1.7.1 ↗ kubernetes/autoscaler
- [2] vertical-pod-autoscaler-chart-0.11.0 ↗ kubernetes/autoscaler
- [3] fix(discovery): log when a configured custom resource CRD is absent ↗ kubernetes/kube-state-metrics
- [4] feat: add support for label generation for ephemeral volumes in pod ↗ kubernetes/kube-state-metrics
- [5] [pt-br] Replace codenew with code_sample shortcode ↗ kubernetes/website
- [6] [fr] Replace codenew with code_sample shortcode ↗ kubernetes/website