• Fix a number of issues pulling China region node prices on AWS
• More accurate pricing for AWS EBS CSI drivers
• Fix CORS and other installation path issues on open source
• Migrate to new Allocation view. Legacy view to remain available at /detail.html
  • The 1d "cache warming" on /aggregateCostModel API is now disabled by default. We will be gradually moving away from this api to the allocation api for scaling and consistency. Most teams will see a dramatic decrease in prometheus/thanos/cortex usage using the allocations API -- #865
• Swap default algorithm for request sizing over 7d and 30d to take “max” usage instead of “p99”
  • This supports getting data for request-sizing over 7d+ windows on larger clusters.
• Alerts migration:
• Note that our diagnostics alert has migrated to use the same block as all other alerts. More info here: https://github.com/kubecost/docs/blob/bolt/alerts-1.79.0/alerts.md#type-kubecost-health-diagnostic
• Deprecate cost-analyzer-checks. This functionality has moved to within the cost-model pod, so there is no reason to run cronjobs anymore. We are not issuing a cost-analyzer-checks-pod at v1.79.0— by default we have set the value to disabled in the helm chart.
• Allow one-click rebuilds of failed cache builds
• Upgrade default thanos installation to 0.19
  • This will have major performance and stability improvements for users installing our bundled Thanos.
  • We have noticed in a percentage of upgrades, the v0.15 thanos-compact will not successfully terminate. It is safe to kubectl delete the thanos-compact pod and any thanos-compact pods in the Pending state to complete the upgrade.
  • You will also need to trigger a rebuild of our cache manually. You can do this by visiting:
    • /model/etl/assets/rebuild?window=all&commit=true
    • followed by /model/etl/assets/rebuild?window=all&commit=true