Restart one service's steps

The service's type in harthelm.yml; steps, container, forward or managed.

Service plan id (steps and container services).

What the spec asks for.

What runs; 0 after a teardown and while asleep.

started: the steps run, also after a pod restart or a wake; stopped: the pods stay, the steps do not run.

• pending: no replica is scheduled yet; starting: replicas start or run their first steps;
• running: the last step runs and the health check passes (internal instead for a service without routes or public ports);
• finished: the last step exited 0 ("run finished"), or a container's process exited 0; failure: a step failed, or the last step (a container's process) failed 5 times in 2 minutes. Both come from the agent's stage result (run_result), not from the pod;
• stopped: the run switch is off.

The most replicas PATCH …/services/{svc} accepts (maxReplicas, else the plan's); steps and container services.

The image as harthelm.yml names it (container services, and steps services with a base image of their own).

Container services, the image as the last Sync resolved it, <repository>@sha256:<digest>; what the pods run.

Container services with runAsUser (teams a cluster admin allowed); absent, the workspace's UID.

The service runs without the run stage (container services): its process starts with its pods and stops with the service's switch; Run and Stop of the landscape leave it alone.

Forward services, the resolved target URL.

Forward services, the paths of the dev domain that go to target.

The last stage result an agent reported for the service: the end of its run steps (steps), or of its process (container). status is failure or finished from it while generation is the service's current run generation.

The service's pods run a template from before an upgrade until it sleeps or is synced again.

Managed services, the instance the Sync created (GET /managed-services/{ms}).

• creating, updating, pausing, unpausing, deleting: the backend works on a change;
• synchronized: it runs as asked; paused: data kept, no server (a torn-down landscape's database);
• deleted: a tombstone, listed with include_deleted until it is purged;
• unknown: the backend does not answer or does not report it; the details are the last known ones;
• invalid-provider: its provider version is gone or no longer accepts its settings.

The services it needs as harthelm.yml lists them, in name order; waking it wakes them, and Run starts them first. Absent when it needs none and from an older core.

0.10.1: the service fails for the services that need it, as failing_needs says it: its run failed, it has no ready replica a minute after its wake, or a Run began while it failed and has not brought up a ready replica yet. A service that needs it neither waits for it at a wake nor at a Run, and shows "needs …, which is failing". Computed when read; absent when false and from an older core.