Performance/Scalability
- Thanos performance improvements related to memory usage and heavy traffic scenarios via upgrade to 0.22.0
- Improved performance and simplicity of PV cost group-right query
- KSM metric emission refactor, improving both cpu and memory performance
- Cloud integrations refactored to their own pipeline
Cost Model
- Option added to control whether or not network reconciliation is done when there is no allocation data and no proportional split. (defaults to true, which does reconcile data)
- Provide consistent experience for pod allocation when only the controller name changes
- cadvisor compatibility improvement now supports either container or container_name
- Additional guardrails added to address erroneous data points received from certain Windows cadvisor installations
Improved testing for containers/pods without a controller - Fix for overreporting of GPU requests for VGPUs
- Additional filtering added to omit metrics for unavailable and unbound PVs
Configuration (from the community!)
- Provide the ability to add extra volumes to the cost-analyzer pod
Frontend
- Update to ensure all html accessibility is controlled via SAML
- Added link to our Troubleshooting Guide
- Cloud status types added for better diagnostics
- Fix for label matching issue related to Allocation aggregation after label update
- Fix for aggregation issue where defaults were being used instead of configured labels
Documentation
- AWS Marketplace installation documentation updates
- Removed reference to helm2 from installation documentation