v2.5.0-rc.4

Major Features

• GPU Savings
• Turbonomics Integration

Minor Features

• UI Consistency with date pickers
• Updated logos and branding to reflect IBM acquisition.
• Add GPU Max usage and GPU Sharing collection to agent, and output to allocation API.

Fixes

• Fix a nil panic on recommendations endpoint.
• Fix an issue with assets labels for the network asset type are showing node labels that are non-network related.
• Update dependencies from go-migrate-duckdb library to fix CVE.
• Fix UI empty state for tables across the user interface.
• Fix an issue allowing for negative values in allocations.
• Fix an issue with accumulation on hourly windows in allocations.
• Fix an issue allowing requestSizingV2 to return more than one recommendation.
• Fix an issue with sharing by namespace distributing incorrectly.
• Fix an issue with allocations trends inaccuracy.
• Fix an issue with the overview not always matching the allocations page for idle.
• Fix an issue causing agent to get HTTP 500 error on /metrics endpoint with duplicate label/annotation.
• Fix an issue where container information was not showing on the allocations page.
• Fix an issue causing oracle clusters to get a crash loop when installing the agent.
• Fix an issue with accounts not always showing correctly in Assets view.
• Fix cost metadata to be more accurate for oracle cloud provider.
• Update kubectl cost to mitigate vulnerabilities in dependencies.
• Fix an issue causing HTTP 500 in multi-cluster diagnostics.
• Fix log level of critical object store outage in the agent to always be error instead of certain conditions allowing warning.
• Fix an issue with ‘uh oh’ showing on allocations page due to an error in /topline endpoint.
• Fix an issue with Assets autocompletion when label filter is set.
• Fix allocation summary efficiency not matching full allocations api.
• Fix an issue where allocations totals row shows shared costs when sharing is disabled.
• Fix an issue with datadog plugin error on database monitoring costs.
• Fix an issue with cloud reports not saving properly.
• Fix an issue with budgets “view breakdown” not matching export output.
• Fix an issue with container right-sizing CPU/Memory columns using inconsistent units.
• Fix an issue filtering by project on the assets page.
• Fix an issue with Multi-cluster diagnostics returning false positives when SSO is enabled.
• Fix an issue with slow response on allocation api with no aggregate set.
• Fix an issue with inconsistent drill-down in cloud costs.

Helm Changes

• #3605 SetupInstanceAllowLists.
• #3591 Add ingestion configuration for standard discount.
• #3635 Add routes for savings recommendations allow list validation API.
• #3647 Add new tmp directory for new base image requirements on frontend container.
• #3645 Add resource reference files for kubecostProductConfigs.savingsRecommendationsAllowLists.
• #3651 Increase prometheus.server.retention to 97h.
• #3658 Remove helm rollout restarter.
• #3642 Add new container costs and resources endpoints to nginx.
• #3660 Enhance Pod utilization Grafana dashboard with GPU Utilization widget.
• #3670 Add aggregator custom labels template.
• #3673 Add federated storage warning.
• #3563 Update version compatibility matrix.
• #3534 Add node selector and affinity to cluster controller.
• #3680 Begins CI testing with K8s 1.31. Bumps patches for all other versions in the matrix. Rolls forward for compatibility of 2.4 by one version.
• #3684 Add /diagnostic/nodeCount endpoint to nginx routes.
• #3622 Make basic-health image configurable.
• #3585 Add missing container security context in prometheus server statefulset.
• #3679 Allow defining resources for cluster controller.
• #3676 Add option for additional chart labels in values.
• #3678 Add annotations to all controllers.
• #3704 Add gpuUtilization endpoints to nginx routes.
• #3708 Add support for multiple simultaneous plugins.
• #3711 Add nginx routes for /savings/gpuUtilization/topline endpoint.
• #3710 Add cost model image tag as an environment variable.
• #3714 Update ISSUE_Guidelines to redirect to https://support.kubecost.com
• #3717 Add logLevel as a helm value property of kubecostModel instead of extraEnv.
• #3716 Fix proper handling of a list of values for imagePullSecrets.
• #3700 Add kubeRBACProxy property to support querying in cluster prometheus in openshift.
• #3694 Delete many deprecated templates/values.
• #3719 Add nginx routes for gpu recommendations endpoints.
• #3723 Add additional aggregation options to endpoint configuration.
• #3725 Update comments about OIDC RBAC groups.
• #3706 Add turbonomic secret.
• #3728 Add nginx routes for topline endpoint.
• #3735 Improve comments in extraScrapeConfigs.
• #3739 Update endpoint configuration for savings/gpuWorkloadUtilization.

Helm Fixes

• #3621 Use ghcr.io instead of dockerhub due to rate limiting for k8s-sidecar
• #3637 Fix diagnostics and federatedStorageConfig
• #3732 Fix turbonomic integration environment variables.
• #3741 Fix upstream chainguard dependency failures.
• #3742 Fix link to docs.kubecost.com.
• #3721 Fix openshift values to default with in-cluster prometheus.

Dependency Updates

• #3724 Bump kubecost modeling 0.1.18
• #3621 k8s-sidecar moved to ghcr instead of dockerhub
• #3662 Bump network-costs to 0.17.6
• #3652 Bump Cluster Controller 0.16.9
• #3741 move all chainguard references back to default upstream image locations.

Helm Chart Comparison Report kubecost/[email protected] to kubecost/[email protected]

CVE by Severity

Severity	Count	Prev Count	Difference
critical	0	0	+0
high	0	0	+0
medium	14	16	-2
low	66	70	-4

Images

Image Name	Status	Before Repo	After Repo	Before Tag	After Tag
kubecost-modeling	Changed	gcr.io/kubecost1	gcr.io/kubecost1	v0.1.16	v0.1.18
cost-model	Changed	gcr.io/kubecost1	gcr.io/kubecost1	prod-2.4.3	prod-2.5.0-rc.4
frontend	Changed	gcr.io/kubecost1	gcr.io/kubecost1	prod-2.4.3	prod-2.5.0-rc.4
prometheus	Unchanged	quay.io/prometheus	quay.io/prometheus	v2.55.1	v2.55.1
k8s	Unchanged	alpine	alpine	1.26.9	1.26.9
k8s-sidecar	Unchanged	ghcr.io/kiwigrid	ghcr.io/kiwigrid	1.28.0	1.28.0
grafana	Unchanged	grafana	grafana	11.3.1	11.3.1

Unchanged CVEs

CVE ID	Severity	Affected Images

Medium

CVE ID	Severity	Affected Images
CVE-2021-3997	medium	cost-model
CVE-2023-30571	medium	cost-model
CVE-2024-35325	medium	cost-model
CVE-2024-50602	medium	k8s-sidecar
CVE-2024-52533	medium	cost-model
CVE-2024-9287	medium	kubecost-modeling

Low

CVE ID	Severity	Affected Images
CVE-2019-12900	low	cost-model
CVE-2022-27943	low	cost-model
CVE-2022-29458	low	cost-model
CVE-2022-3219	low	cost-model
CVE-2022-41409	low	cost-model
CVE-2022-4899	low	cost-model
CVE-2023-2953	low	cost-model
CVE-2023-32636	low	cost-model
CVE-2023-36191	low	cost-model
CVE-2023-4156	low	cost-model
CVE-2023-45322	low	cost-model
CVE-2023-45918	low	cost-model
CVE-2023-50495	low	cost-model
CVE-2024-0232	low	cost-model
CVE-2024-2511	low	cost-model
CVE-2024-34459	low	cost-model
CVE-2024-41996	low	cost-model
CVE-2024-4741	low	cost-model
CVE-2024-51744	low	grafana
CVE-2024-5535	low	cost-model
CVE-2024-7264	low	cost-model
CVE-2024-9143	low	grafana, k8s-sidecar
CVE-2024-9681	low	cost-model

Added CVEs

No CVEs found.

Removed CVEs

CVE ID	Severity	Affected Images

Medium

CVE ID	Severity	Affected Images
CVE-2024-49766	medium	kubecost-modeling
CVE-2024-49767	medium	kubecost-modeling

Low

CVE ID	Severity	Affected Images
CVE-2024-51744	low	cost-model
CVE-2024-9143	low	kubecost-modeling

v2.5.0-rc.3

• #3721 Fix openshift values to default with in-cluster prometheus.
• Fix docs link referencing guide.kubecost.
• Add savings/gpuWorkloadUtilization endpoint to nginx config.

v2.5.0-rc.2

• Continue to iterate GPU Savings with bug fixes and enhancements. GPU Sharing and MaxUsage is now available.
• Fix an issue with multicluster diagnostics not reporting correctly.
• Improved OIDC handler logging.
• Add minutes to SingleAllocationSummary query.
• Only predict node costs if provider exists in cloud costs.
• Fix issues with cost sharing in allocations where sharing could cause total costs to increase inadvertently.
• Fix issue with turbonomics integration causing ingestion to show as not completed.
• Fix an issue with request sizing allowing multiple recommendations to be returned in the API.

v2.4.3

Fixes

• Resolve prometheus image pull error.
• Resolve issue with duplicate annotations causing 500 in kubecost agent metrics endpoint.
• Resolve issue with overview page gpu/cpu/ram efficiency on large datasets.

v2.4.3-rc.0

Resolve issues pulling prometheus container image.

v2.5.0-rc.1

• Fixes several UI issues in GPU Savings experience

V2.5.0-rc.0

• GPU Savings
• Turbonomics Integration

v2.4.2

What's Changed

• Reduced aggregator memory footprint during derivation
• Add Oracle Cloud allocation support to Kubecost
• Restore logout button for SSO
• UI scrolling does not work after closing cloud cost modal
• Improve performance on certain allocation API calls
• Add missing container security context in Prometheus server statefulset #3691
• Add cluster controller resource helm template #3692

Full Changelog: v2.4.1...v2.4.2

v2.4.2-rc.0

Fixes

• Allow allocations to be created from oracle cloud oke clusters.
• Allow node group sizing to use same-family recommendations.

v2.4.1

Fixes
#3685 Fix issue with invalid column name after database upgrade.
Fix an issue with idle weighted by cluster after new GPU Idle was added.
Fix an issue with request-sizing page, ensure compatible query window is used.
Fix an issue showing an error on cloud costs page inadvertently.
Fix an issue where topline allocations costs changes depending on idle and shared cost settings.
Helm Fixes
#3674 When a federated store is provided, Kubecost should either be running as agentOnly or as a statefulset.
#3684 Add /diagnostic/nodeCount endpoint to helm chart.