Observability
Added
- Add labels to Deployment, DaemonSet, StatefulSet metrics:
app.kubernetes.io/version, helm.toolkit.fluxcd.io/name, helm.toolkit.fluxcd.io/namespace
Changed
- Upgrade Tempo chart from to 1.61.3 to 2.4.2
- Upgrades Tempo from 2.9.0 to 2.10.1
- Upgrade Tempo Vulture chart from 0.12.5 to 0.12.6
Changed
- Update kube-prometheus-stack to 20.1.0
Added
- Add Gateway API
HTTPRoute resources for Loki, Mimir, and Tempo (read and write), replacing the previous NGINX ingress setup. - Add native JWT authentication via Envoy Gateway
SecurityPolicy.jwt, supporting multiple OIDC providers (e.g. Dex, Azure AD). Configurable via auth.jwt.providers. - Add
/loki/api/v1/rules to the Loki read routes. - Add
GRPCRoute for Tempo gRPC traffic (port 9095), routing all tempopb.* services to tempo-query-frontend with JWT enforcement via SecurityPolicy.
Changed
- Replace NGINX ingress-based auth (
nginx.ingress.kubernetes.io/auth-url) with Envoy Gateway SecurityPolicy JWT validation — no external auth service (oauth2-proxy or Dex extAuth) required. - Change missing
X-Scope-OrgID response code from 400 to 401 across all routes. - When
auth.jwt.providers is empty and a service is enabled, routes are silently not rendered (no chart error). Previously the chart would fail with an error. - Fix Tempo gRPC route service regex from
tempopb to tempopb\.[^/]+ to correctly match package-qualified service names (e.g. tempopb.StreamingQuerier).
Removed
- Remove dependency on oauth2-proxy for write route authentication.
- Remove Envoy Gateway
Backend CRD and extAuth configuration in favour of inline JWT validation.
Changed
- Upgraded chart dependency to kube-prometheus-stack-82.8.0
- Added VPA support for Alertmanager
- Added VPA support for Prometheus
- grafana from 11.2.2 to 11.2.3
Changed
- Upgraded chart dependency to kube-prometheus-stack-82.8.0
- Added VPA support for Alertmanager
- Added VPA support for Prometheus
- grafana from 11.2.2 to 11.2.3
Added
- Add KSM metrics for Envoy Gateway resources.
- Add
application.giantswarm.io/team annotation from HelmReleases as label to KSM emitted metrics.
Changed
- Change team annotation in
Chart.yaml to OpenContainers format (io.giantswarm.application.team). - Update alloy-app to 0.17.1
- Update kube-prometheus-stack to 20.0.0
- Update prometheus-operator-crd to 20.0.0
Changed
- Upgrade Alloy upstream chart from 1.6.0 to 1.6.1 (CHANGELOG)
- This bumps the version of Alloy from 1.13.0 to 1.13.2 (CHANGELOG)
Fixed
- Fix memory usage calculation in
nodes-overview and cluster-overview dashboards by using node_memory_MemAvailable_bytes instead of node_memory_MemFree_bytes, which incorrectly excluded cached/buffered memory from free memory
Changed
- Update DNS dashboard
- Add new node and pod filters
- Update variables description and query to use coredns_build_info as label source
- Remove cache prefetch panel since the metric is gone
- Fix DNS dashboard log panels
Kube-Builder Operators dashboard: add a logs datasource selector