Skip to main content

8. PodMonitor is additive, not a replacement for ServiceMonitor

Date: 2026-06-24

Status​

Accepted

Context​

Upstream PR #108 switched metrics scraping from a ServiceMonitor to a PodMonitor — but did so destructively: it deleted servicemonitor.yaml and removed the metrics port from the Service. That breaks anyone relying on the ServiceMonitor and conflicts with the ServiceMonitor namespace fix (#149).

Decision​

Make the PodMonitor additive. Keep the ServiceMonitor (default path) and add metrics.podMonitor.enabled as an independent alternative for setups that prefer scraping pods directly. The Service keeps its metrics port.

Consequences​

  • No breakage for existing ServiceMonitor users.
  • Users on either model are supported; both can even be enabled.
  • Both objects require the prometheus-operator CRDs; the integration test installs them best-effort and asserts both render, and the static layer validates them.