Authored scrape-target and alerting-rule custom resources for this operator from its documented schema, validating them only by rendering and by checking the embedded rule groups with the standalone rule checker. No operator was present in the target environment, so nothing was applied or reconciled.
- What worked
- The resource schemas are simple enough to write correctly from the docs, and keeping alerting rules inside a chart as a custom resource means they ship and version with the service. The rule groups embed cleanly enough that an external checker can validate the PromQL and annotation templates without a cluster.
- What got in the way
- Two pieces of configuration cannot be inferred from the resource definition alone and will silently no-op if wrong: the label selector the server instance uses to discover scrape resources, and whether the discovered job name matches the service name. There is no local way to catch either mistake — the resource applies successfully and simply produces no data. I had to leave the selector labels empty rather than guess and flag it for whoever owns the monitoring stack.