This page describes Darling v3.13.0. The source of this page is on GitHub at that version.
Audience: operators.
The repository carries a Helm chart for Darling: Darling/helm/performancemonitor-darling. It runs the service as one pod with a state volume, with either a bundled TimescaleDB pod or a connection to a PostgreSQL server you run. The chart is not published to a registry yet, so you install it from a checkout:
helm install darling Darling/helm/performancemonitor-darling --set-string image.tag=<release>
Pin image.tag to a release, because latest moves. Use --set-string, not --set, because a float-looking tag such as 3.10 loses digits otherwise.
The chart README is the reference. It covers:
- the two store modes, and why custom alerts need a managed or compose store for now;
- which values the store overrides after the first start (see also
darling.jsonis only a seed); - secrets, and why a GitOps install must name its own Secrets;
- the password key (see also Keeping the key outside the service);
- readiness, which follows the collector (see the health port);
- upgrades;
- uninstall, which keeps the volumes and the Secrets.
Out of the box only kubectl port-forward reaches the dashboard. An Ingress needs allowFrom per surface (the dashboard judges the client address the controller forwards, the MCP endpoint the controller’s own connection) and, for the dashboard, trustedProxies; see reaching it from outside the cluster and Opt-in Network Endpoints.
If you write the manifests yourself instead, the pod needs the securityContext and the volume ownership steps in Container user.