Fixed

• Add split idle back to allocations api.
• Fix issues with network cost reconciliation for Azure providers.
• /savings/requestSizingV2 endpoint will now work for clusters containing nodes without pods.
• Fix an issue gating EKS Optimized assets query unnecessarily.
• Fix an issue where allocation ‘other’ cost discrepancy when idle is included.
• Fix a nil panic on cluster sizing recommendation.
• Fix 500 error on cluster sizing when provider is missing with incorrect configuration.
• Fix issue with teams showing test data in certain configurations.
• Fix the service accounts learn more button to correctly link to documentation.
• Fix RBAC configuration check in the frontend.
• Fix shared cost modal to properly display sharing idle, tenancy costs, and fixed costs.
• Fix allocation page to support aggregations to specific controller kinds.
• Fix assets top line totals not matching table correctly.

Helm Changes

• #3303 Helm value changes for external-costs, now off by default because it requires configuration
• #3297 Fixed Grafana service account creation when Kubecost service account creation is disabled
• #3292 Added extraVolumeMounts to cloud-cost and aggregator templates
• #3321 Adds security context values for cloud-costs when openshift is enabled.
• #3334 Add missing read_only environment variable.
• #3335 Set the daily retention limit in the cloud-cost container which allows it to build historic cloud data to the correct retention limit.
• #3325 Fix kubecostFrontend.deployMethod default value.
• #3330 Fix an issue allowing for cloud cost to run in Opencost Cost model when single pod deployment method is chosen and a cloud cost container is deployed as well.

Known issues (coming in 2.2.3)

• Network Monitor is not working in single cluster environments
• Abandoned workloads threshold and duration are broken

Fixed

• Custom cost plugin installation fixed.
• Changes to the process of ingesting data. Processing windows in descending order across clusters to get the most relevant information visible compared to whole clusters at a time.
• Fix Scheduled reports bug stopping reports from sending.
• Fix real-time estimation of node prices causing large inaccuracies in allocations for unreconciled data.
• Fix broken link to documentation for cloud integration.
• Fix an issue where Cloud Costs page loaded slowly without External Costs configured
• Update domain links to docs links in front-end from docs.kubecost.com to gitbook.kubecost.com.
• Add a diagnostic message to the Top Traffic Destinations card on Overview to explain when network-costs are enabled, but data isn’t being properly received.
• Fix an issue where Overview’s top traffic destinations did not link out correctly.
• Fix an issue where the Savings page’s card for Abandoned Workloads savings was vastly different from the number on the Abandoned Workloads page.
• Fix an issue where the version hash of the UI was not printing to the settings page.
• #3306 Fix CVE in kubecost-modeling container for CVE-2024-3094 xz vulnerability.

Helm Changes

• Helm value changes for external-costs, now off by default. Users need to configure integration before turning this feature on.
• Fixed Grafana service account creation when Kubecost service account creation is disabled.
• Bump kubecost-modeling image to 0.1.6 to fix CVE-2024-3094 xz vulnerability.

Overview: Kubecost v2.2 is considered an edge release with many new features:

Major

• New External Costs page provides cost monitoring for third party services. Currently supports plugin integration for Datadog costs
• New cost metric for Carbon emissions equivalent added to Allocation and Assets pages
• Simplified multi-cluster Prometheus (AMP/Thanos/GMP/Mimir) without the need for shared object-storage

Minor

• Enabled feature gates that Kubecost docs have indicated were Enterprise only, but not enforced - this includes some savings related to multi-cluster and eks-optimized.
• Enhanced diagnostics on clusters overview page (more to come)
• Collections now supports exporting to CSV and PDF formats
• Forecasting improvements for Allocations, Assets, and Cloud Costs
• Enabled authentication with EKS Pod Identities
• Enabled authentication with Azure Workload Identities support with Federated ETL
• Added UI status indicator to display if backups are configured
• [Enterprise Only] HA Mode - phase 1: Frontend always available during restarts, UI status indicator
• [Enterprise Only] Multi-cluster cluster right-sizing improvements with added support for high scale environments
• Added support for filter-based continuous right-sizing
• Added quantile support for Request Right-Sizing Recommendations page
• Added pagination support to the Collections page
• Fixed broken links in UI to product documentation
• Reduced default Prometheus retention from 15 days to 97 hours for stability

Bugs

• Fixed issues with shared cost breakdown
• Fixed the ability to use hourly data for request-sizing based on window availability
• Fixed onDemand pricing by node
• Fixed issues with kubectl cost predict working with new v2.0 Aggregator architecture
• Changed values.yaml tolerations properties from {} to []
• Update frontend /healthz endpoint to be accessible with OIDC installations
• Cleaned up services created in agent-only configuration
• AllowsLOG_LEVEL to be configured for network costs daemonset
• Removed free trial days remaining count in UI when using an Enterprise license
• Fixed an error which prevented the /allocation/trends API from functioning.
• Fixed a bug that broke Kubecost UI for SAML users after helm upgrades

Dependency changes

• Updated kubecost/prometheus-config-reloader image from v0.71.2 to v0.72.0
• Updated kiwigrid/k8s-sidecar image from 1.25.4 to 1.26.0
• Updated grafana/grafana image from 10.3.1 to 10.3.4
• Updated prometheus/prometheus from v2.49.1 to v2.50.1
• Updated prometheus/alertmanager from v0.26.0 to v0.27.0
• Updated gcr.io/kubecost1/kubecost-modeling from v0.1.3 to v0.1.5
• Added image curlimages/curl:latest when using externalCosts plugin

Bugs Fixed

• Fix an issue causing multiple shared labels with the same key not sharing.
• Fix an issue with saml users causing issues on allocations filters.
• Fix an issue in cluster-sizing pricing.
• Fix several issues with allocations csv and pdf exports.
• Fix an issue to show an appropriate error when coreCount endpoint isn't available.
• Fix an issue with cloud account mapping.
• Fix an issue with monthly step size in allocations.
• Fix an issue with hourly cost calculations in allocation detail drilldown.
• Fix an issue with teams delete team member.
• Fix an issue with teams saving with OR filters.
• Fix an issue with cluster profile.

Security Updates

• Bump dependency jwx to 1.2.28 for security improvements GHSA-7f9x-gw85-8grf

Helm Updates

• #3176 Fix failed install/template with no values on older versions of Helm3
• #3180 Correct image for networkCosts image in default values
• #3208 Add new LOG_LEVEL to network costs, defaulted to info

Important Notice

If you are upgrading from any version less than 2.0 (Eg. 1.108 or prior) - You will lose your saved reports on upgrade if you upgrade directly to 2.0 - We are currently working on a workaround to copy reports to your new install! We'll post an update here when we have it!

Major

• Service Account Access
• Improvements to the ingestion process to allow for incremental derivation
• Pricing/Packaging updates to core based

Minor

• Multi-cluster Diagnostics
• Customer Success Report
• Performance Enhancements to Collections Feature
• Add OIDC support
• YAML Configuration for Kubecost Actions
• Improve 90+ day queries by setting the default step size to weekly
• Displays an error message when cloud service data doesn’t load
• Anomaly Detection sorts by cost as default, and adds client-side searching
• Cloud Cost Explorer now supports filtering by custom label
• Updates made to default step sizing
• DurableStorage diagnostics removed from Settings page
• Kubecost’s Aggregator API endpoints are unavailable to non-Enterprise users with more than 200 total cores in their environment over the last 30 days
• Added support for multi-cluster Prometheus tools (AMP/GMP/Grafana cloud/Mimir)
• Cluster Controller image updated to v0.15.0
• Added skipSanityChecksvalue which when enabled will prevent Helm from performing sanity checks for cloud integration secrets

Bug fixes

• Fixed a bug where the Allocations page would crash when trend percentage values wouldn’t load properly
• Fixed a bug in Collections where there was inaccurate overlap between Kubernetes and Cloud costs
• Fixed an issue where network costs for certain Allocation data could be large negative numbers
• cloud-integration.json file now contains athenaWorkgroup and masterPayerARN values for AWS cloud integrations

Important Notice

If you are upgrading from any version less than 2.0 (Eg. 1.108 or prior) - You will lose your saved reports on upgrade if you upgrade directly to 2.0 - We are currently working on a workaround to copy reports to your new install! We'll post an update here when we have it!

Minor

• Now showing a more informative empty state on Anomaly Detection when a cloud provider is not integrated.
• Now showing a more informative error state on Forecasting graphs, depending on the nature of the failure (aggregation not supported for forecasting, model failed to build, or request was malformed).Bugs
• #49 Fix an issue causing ResolveEndpointV2 error when pulling AWS Cloud * Costs
• Fix an issue with ingestion of cost data on larger data sets
• Fix an issue where the sub-menu to set custom start/end dates would render behind the main window selection menu.
• Fix an issue where viewing the details of a team immediately after creating the team opened an empty detail panel until the page was refreshed.
• Fix a broken docs link in the new Networking view.
• Fixes a bug where cloud billing via KubecostProductConfigs was not honored.

Hey there! We wanted to say thanks for all the feedback on our 2.0.0 release today! We got some incredible feedback and wanted to address it quickly. Please keep it coming! Drop Github issues or email us at team@kubecost.com. We’re committed to making sure we address all of the feedback we get as quickly as possible. Thank you!

Important Notice

If you are upgrading from any version less than 2.0 (Eg. 1.108 or prior) - You will lose your saved reports on upgrade if you upgrade directly to 2.0 - We are currently working on a workaround to copy reports to your new install! We'll post an update here when we have it!

Minor

• Add a confirmation dialog before deleting a Collection.
• Show a friendlier error message when attempting to generate a cost forecast for an unsupported aggregation.

Bug Fixes

• Fix aws cloud cost connection issues.
• Fix a consistent console log error.
• Fix an issue where bar charts in Collections would have dates off-by-one on the x-axis.
• Fix an issue where a message informing users that they must have SSO enabled in order to work with Teams was always hidden.
• Fix an issue on the Overview’s top traffic destinations table that always shows 0 bytes transferred.
• Disabled forecasting pod when agentOnly is enabled.
• Fix an issue with aggregating by label in Collections.
• Fix an issue with persistent volume savings page loading properly.
• Fix an issue aggregating cloud cost by label for aws cloud costs.
• Fix an issue on the providers page due to cloud cost status not returning appropriate data.
• Fix an issue with azure network reconciliation causing inflated network costs on azure workloads.

Kubecost 2.0 Release Notes

Announcing Kubecost 2.0!

Important Information:
Kubecost 2.0 is a major upgrade from previous versions and includes major new features including a brand new API Backend. Please review the following documentation to ensure a smooth transition: https://docs.kubecost.com/install-and-configure/install/kubecostv2
For the full list of enhancements, please see our release notes: https://github.com/kubecost/cost-analyzer-helm-chart/releases/tag/v2.0.0

Important Notice

If you are upgrading from any version less than 2.0 (Eg. 1.108 or prior) - You will lose your saved reports on upgrade if you upgrade directly to 2.0 - We are currently working on a workaround to copy reports to your new install! We'll post an update here when we have it!

Major

• Kubecost Aggregator becomes part of Kubecost’s default architecture and API backend, replacing Federator component of federated ETL environments
• Thanos, query service replicas, and cost-model statefulSet mode are no longer supported as Federated ETL + Aggregator becomes default. Kubecost customers using Thanos for multi-cluster metric federation should consult the migration guide to ensure a seamless upgrade and retain all historical data access
• New Advanced Networking page visualizes network connections and displays ingress/egress costs
• Kubecost Actions promoted to GA out of beta: UI updates to Kubecost Actions page and features Automated Request Sizing, Cluster Turndown, and Namespace Turndown
• New Collections page to replace Cost Center Reports
• New Anomaly Detection page displays potential irregular spend in cloud services
• New Teams page which supports assigning RBAC to users in Kubecost environment
• Visualization added to Container Request Right-Sizing page displaying requested vs. recommended spend; additional filtering and sorting added
• Forecasting feature across Allocations, Assets, and Cloud Cost Explorer pages which displays projected future costs
• Label tags added to Cloud Cost Explorer line items for additional filtering

Minor

• Filter button added to Allocations, Assets, and Cloud Cost Explorer dashboards, and Persistent Volume Right-Sizing Recommendations page
• Failure message appears when trying to install/upgrade Kubecost with pod security policies enabled
• Users of the Aggregator Beta releases (available in v1.107.X and v1.108.X of the Helm chart) will need to delete and recreate the Aggregator StatefulSet before upgrading to v2.X because the Volume configuration has been altered. This will initiate a rebuild of the Aggregator data store; expect some downtime upon upgrading. No data will be lost.
• Updated Cluster Controller to v0.14.0
• Updated default row count on pages for Allocations, Assets, and Cloud Cost Explorer from 10 to 25.
• Fixed a bug where Network Costs did not display any cross-zone traffic
• Updated to current versions of Kubecost Cluster Controller, Grafana, Prometheus, push-gateway, configmapReloader, alertmanager
• Node-exporter and Prometheus configMap Reloader are now disabled by default
• If configMap reloader is enabled, fix duplicate security context
• Add coreCount endpoint
• Add ingress className for prometheus server
• Removed helm subcharts, moved required components to main chart
• Liveness probes check from 30s to 10s, faster pod startup
• Fix pod probe templates

Known Issues

• Sharing parameters of the /model/allocation API have no effect in v2.0.0. If you consume the /model/allocation API directly and need sharing support, please consider if you can use the /model/allocation/summary API as a stopgap or wait until an upcoming release reintroduces support. Sharing via Kubecost's web frontend is unaffected and behaves as expected.
• The Inspect Shared Costs button on the Allocations page of Kubecost's UI will not display information. An upcoming release will reintroduce support for this feature.
• Adds a section to bug reports for multi-cluster diagnostics

Bugs

• Fixed a bug where drilling down into assets provider item attempted to filter by providerID
• Fixed a bug where the Allocation Details page did not show the names of cloud services linked to a namespace when data was derived from the Cloud Usage pipeline.
• Fixed parsing error which occurs after removing time zone from timestamp fields
• Fixed a bug where daily step sizes would fail to load Allocations data
  And many, many, many...many more

Bug Fixes

• Fix an issue in allocation inspect page to properly display cloud cost service name in table.

Bug fixes

• Fix an issue where the UI would hide diagnostic elements whether or not the flag to do so was enabled.
• Fix an issue with totals in the Assets table being incorrect.
• Fix an issue where filters like "Contains" and "Starts with" could not be used, because the completion Combobox did not allow submitting partial text.
• Fix an issue where the Request Sizing results table could not be paged in non-aggregator installs.
• Fix an issue where a cluster's "Last Seen" reporting status would be in the future.
• Added missing GCP resource discounts
• Fix an issue where nginx would run as root. Ports under 1024 are no longer supported.
• Removed unused telemetry metrics from opencost
• Add systemProxy to cloud costs
• Add helm templates to readinessProbe and livenessProbe
• Reduce default probes to 10s from 30s
• Fix an issue where cluster-sizing in aggregator would not have any results
• Removed init container from query service replicas