Expand description
Admin HTTP server: /metrics, /healthz, and /readyz.
One small hyper server per process, running on the I/O runtime. Probe
semantics (see docs/DESIGN.md § Errors, panics, shutdown, health):
/readyz— 200 once the source has received its assignment and the sinks are connected./healthz— 200 while every pipeline thread’s poll-loop heartbeat is fresh AND the checkpoint watermark is not stuck while data flows. A stalled poll loop or a wedged watermark flips this to 503 so the kubelet restarts the pod.
The server takes the metrics renderer as a plain closure
(RenderFn) so it has no dependency on exporter internals.
Structs§
- Admin
Server - The admin server, bound and ready to run.
- Health
State - Shared health state, updated by the pipeline runtime and read by the probes. All methods are lock-free and callable from any thread.
- Health
Thresholds - Staleness thresholds for the liveness probe.
Type Aliases§
- Render
Fn - Renders the current metrics exposition snapshot.