1. Docs
  2. Changes and releases
  3. Observability

Observability

  • Observability alloy-app v0.23.1

    Added

    • Optional serviceAccountTokenSecret.enabled to render a long-lived service account token Secret.
  • Observability alloy-app v0.23.0

    Changed

    • Upgrade Alloy upstream chart from 1.10.0 to 1.12.1 (CHANGELOG)
      • This bumps the version of Alloy from 1.17.0 to 1.19.2 (CHANGELOG), which includes several CVE fixes.
      • Alloy Breaking changes
        • [prometheus.operator.servicemonitors] ServiceMonitor endpoints referencing bearerTokenFile, tlsConfig.caFile, tlsConfig.certFile or tlsConfig.keyFile are now rejected. Set allow_arbitrary_file_access = true to keep the previous behaviour.
        • [otelcol.receiver.otlp] The HTTP server now defaults to idle_timeout = "1m", read_header_timeout = "1m" and write_timeout = "30s" instead of "0s".
        • See Alloy v1.18.0 and v1.19.0 release notes
      • [beyla.ebpf] Beyla now runs as a subprocess, which raises the memory used by the Alloy container.

    Fixed

    • Allow verticalPodAutoscaler.updatePolicy to be set, including the in-place update modes.
  • Observability dashboards v4.30.0

    Added

    • Add an agent_platform sub-chart (Team Bumblebee) and its first dashboard, Giant Swarm / Agent Platform / LLM usage: tokens, estimated cost and latency for every model call that goes through the Agent Platform’s agentgateway LLM listener, broken down by agent, agent namespace and model. It reads the GenAI metrics the agentgateway data plane emits. Empty until an installation enables llmRouting in the Agent Platform values and points its agents at the listener.
    • Add a Cost accounting row to the LLM usage dashboard, for attributing spend rather than watching it: totals for the selected time range, cost per bucket stacked by agent and by model over a selectable Cost bucket period, and per-model and per-agent-and-model tables with column totals.

    Changed

    • Move the private Falco dashboard to the new Giant Swarm Security organization and rename it to Anomaly Detection, to include other data sources.

    Removed

    • Drop the Security: Falco Dashboard and Falco public dashboard boards, so no Falco dashboards remain outside the Giant Swarm Security organization.
  • Observability alloy-app v0.22.0

    Added

    • Add RBAC to allow alloy-metrics to read the alloy-vcenter-credentials secret when running vSphere or VCD clusters.
    • Add serviceWhenDisabled.enabled, keeping the Service rendered while alloy.enabled is false. Off by default.
  • Observability observability-bundle v3.4.0

    Changed

    • Values: Update Prometheus Operator CRD and Kube Prometheus Stack to v23.0.0.
  • Observability kube-prometheus-stack-app v23.0.0

    Changed

    • Chart: Upgrade kube-prometheus-stack to v88.6.3, including:
      • prometheus-operator to v0.93.1
      • kube-state-metrics to v8.4.1
      • prometheus-node-exporter to v4.56.3
      • grafana to v12.11.2
  • Observability prometheus-operator-app v23.0.0

    Changed

    • Chart: Upgrade kube-prometheus-stack to v88.6.3, including:
      • prometheus-operator to v0.93.1
      • kube-state-metrics to v8.4.1
      • prometheus-node-exporter to v4.56.3
      • grafana to v12.11.2
  • Observability loki-app v0.48.0

    Fixed

    • Keep the provider-kubernetes identity RBAC on uninstall, so the crossplane Objects can drop their finalizers instead of hanging.

    Added

    • Add gatewayRoute.requestMirror, an optional RequestMirror filter on gatewayRoute.additionalRules, plus the ReferenceGrant a cross-namespace backendRef needs. Disabled by default.
  • Observability observability-platform-api v0.5.0

    Added

    • Add loki.write.requestMirror, an optional RequestMirror filter on the Loki write route. Disabled by default.
    • Add opt-in Basic Auth read routes for Mimir and Loki on a dedicated hostname (basicAuth.*, disabled by default).
  • Observability grafana-app v2.40.0

    Added

    • Brand Grafana with the Giant Swarm logo mark and favicon, by mounting the assets in helm/grafana/branding over the ones the Grafana image ships. Set grafana.extraConfigmapMounts to [] to get Grafana’s own branding back.
    • Add make branding-verify (run by the verify-branding CI job) and make branding-update, which keep the hashed logo path in sync with the Grafana image and generate the pod annotation that rolls the deployment when a branding asset changes.

    Changed

    • Upgrade grafana chart: 12.10.3 => 12.11.2
    • Upgrade grafana (appVersion): 13.1.2 => 13.2.0