Important Updates
Kubecost recommends that any environment not using the Kubecost-provided Prometheus instance upgrade all clusters to Kubecost version 1.102+ due to a mismatched label that can cause container costs to be under-reported.
Additionally, Kubecost added a new recording rule in this version to enhance the CPU request-sizing recommendations accuracy when using the max algorithm. For clusters using existing Prometheus instances, this rule can be added by following this guide. Note that even without the recording rule, Kubecost will use a less efficient subquery in order to provide accurate recommendations.
Major
- Ability to query 1 year of Allocations and Assets data
- New Clusters dashboard and improvements to cluster visibility
- Dedicated cluster view page and drilldown
- Added costs for unmonitored clusters
- Show expanded savings info on Cluster Sizing by default
- Add cluster counts to diagnostic alerts for easy identification of missing clusters
- Improved Cloud Costs Explorer features and made progress toward turning on by default
- Added filters to cloud cost
- Added drilldown to cloud cost
- Added multi aggregation to cloud costs
- Improved Audit dashboard experience
- Track statefulset adds and deletes
- Persist events through pod restart
- Reconciliation from Cloud Costs
Minor
- Failed ETL notifications added to UI health alerts and email
- Updated color palette to improve usability
- Improved cloud integration visibility
- Added support for multi-aggregation drilldowns in Allocations
- Added support for weekly step size, storage, and queries for Allocations and Assets
- Most modals will now close when pressing the escape key
- Added additional filters to Allocations dashboard including department, environment, owner, product, team
- Added support for the Indonesian Rupiah (IDR).
- Added HTTP parameter validation to /prediction/speccost. The endpoint will now error if an invalid HTTP parameter is provided.
- Added API for checking the status of the cost analyzer volume
- Displays the PVC storage statistics on the diagnostics page
- Linked to Disk-specific grafana dashboard from health items around PVs and PVCs
- Improved tooltip experience
- Improved navigation
- Allocations dashboard now links to right sizing
- Updated the headers of all Savings pages to navigate back to main Savings page.
Bugs
- In environments that use a non-bundled Prometheus instance: if a pod consumed more CPU and memory than requested, missing labels caused under-reported costs
- Fixed in request-sizing, CPU max algorithm did not properly calculate max
- Fixed an issue where the savings card for Cluster Sizing claimed it was "Cloud Savings," not "Kubernetes Savings"
- Fixed a bug where the Assets dashboard was defaulting to aggregation by Service when there was no cloud integration present (only service is always "Kubernetes"), and defaulting to Type otherwise.
- Fixed bugs causing Cluster Sizing page to crash
- Changed Allocations time series chart tooltips to show dates in UTC instead of local time.
- Fixed an issue where we show errors for pricing source configuration if the user isn't using RIs.
- Resolves libcurl CVE
- Fixed a CVE in golang.ogr/x/net
- Fixed a bug where the reserved instances Savings page would crash when some query results were empty.
- Fixed network cost ingestion for VMSS nodes on Azure
- Fixed reconciliation with EKS Fargate nodes leads to double counting 2096
- Fixed an issue where SAML redirect was using an invalid URL. 2098
- Fixed a bug where drilling into a job wouldn't properly filter/set the correct context.
- Fixed an issue where Savings total was double-counting savings from Unclaimed PVs and not accounting for PV-sizing savings.
- Fixed a bug where links to request sizing from the workload details page would not filter by controller correctly.
- Fixed a bug where the Cluster Sizing Savings card did not leverage the cluster's Profile to show potential savings accurately.