v1.60.0

Application:

• Adopt ETL pipeline: this is the largest architecture change since launch. New ETL pipeline allows building up to 90d allocation views without touching backing datastore, e.g. Prometheus, Thanos, Cortex. Learn more
• Implement new license key store
• Handle missing charts library
• Improve custom provider CSV pipeline
• Better reporting and recovery on panics
• Remove requirement to provide labels for spot nodes on AWS

Helm Chart:

• Fix PSP role name (#506)
• Support GCP keys at install time (#501)
• Support pulling Busybox from any repository
• SAML: Support separate audience URI from app root URL

v1.59.2

• Fix issue with SAML auth failing when .Values.global.grafana.proxy = true. Note that a restart of the cost-analyser pod may be required to refresh the nginx configmap as well as any cookies minted by the cost-analyzer pod

v1.59.1

• Refine Prometheus/Thanos diagnostic tests
• Fix: resize memory measurement kubecost/kubecost-cost-model#142
• Change default GOGC rate
• Reduce Thanos query concurrency
• Add single retry on cache build failure
• Memory optimizations
• Address cluster costs startup stability issue
• ETL: fix panic in accumulation
• Helm: Remove SAML secret requirement
• Helm: Add kubecostDeployment labels option

v1.59.0

Fix concurrency issue when updating pod information from k8s api
Support Thanos as default data source in Grafana
Reduce Prometheus concurrency

• Add PodSecurityPolicies
• Add SAML auth to /api endpoints
• Multiple replica support on kubecost pod
• Fix off-by-one error on AWS Cost Report parsing
• Redact AWS and Azure information on frontend
• Fix PV data duplication if multiple kube-state-metrics running
• Support additional timewindow formats
• Fix config parsing issue with new network cost classifications

v1.58.1

• Fix issue with custom external cloud tags
• Fix log collection nil pointer #452
• Fix serviceaccountname in network costs template
• Add CORS header to abandoned resources endpoints

v1.58.0

• Fix for directly querying labels with "-" in API
• Reporting Flags written to ClusterInfo
• Reserved instance fixes
• WHERE clause to reduce data pulled by Athena queries
• One-click cluster resize
• SAML metadata upload support
• Network costs nil interface fix
• Thanos service discovery support
• v1 cluster turndown released
• Prometheus upgraded to 2.17.2
• Additional prometheus performance tuning

v1.57.2

• Fix issue with reserved instance mapping
• Reduce resource usage of cluster resize PromQL
• Expose warm savings cache flag
• Expose Athena configs at install-time
• Support deep-linking on non-root ingresses

v1.57.1

• Reduce default Prometheus concurrency to address memory concerns in 2.16 release
• Drop Postgres remote_write by default
• Support using an existing service account correctly
• Correctly support running Kubecost on ports less than 1025
• Add gpu label and customPricesEnabled install-time flag
• Move clusterIDConfigmap out of prom global config
• Add matchLabels option to thanos services
• Update Network Costs to v13
• Fix panic when Prometheus is unavailable
• Drop unused ComputeUptimes query

v1.57.0

• Support basic auth for secondary clusters
• Support wildcards for SAML access
• Support custom external cloud cost tags
• Support network whitelist traffic types
• Remove kubernetes-pods and nodes scrape endpoint
• Make grafana URL configurable from values.yaml
• Make cost-analyzer port and targetPort configurable
• Expose cache warming configuration
• Improve load time of Savings page
• Add support for binding IAM roles to thanos on AWS

v1.56.1

• Update Thanos cluster_id diagnostic test
• Update Prometheus cadvisor label diagnostic test