The GitHub Actions job "CI" on rocketmq-dashboard.git/fix/broker-metrics-unavailable-on-fetch-failure has failed. Run started by GitHub user unbridled-41 (triggered by unbridled-41).
Head commit for run: 4f3793d75dbd7c7015cc5ab683b6b96a12cfc2c8 / unbridled-41 <[email protected]> fix(metrics): keep active broker alerts when the runtime stats read fails When a broker's runtime stats cannot be read at all - no master address, a null response, or a failed call - collectBroker recorded only broker.availability and returned, so the disk, JVM-heap and send-queue metrics produced no sample at all. Alert reconciliation reads a missing sample as "the value is gone" and advances an active state with clear: an enabled broker.disk.usage_ratio incident that is FIRING therefore resolves on the very collection pass that failed to read the value, and flaps back when the broker recovers. The alerting design note says a collection failure records an unavailable sample, and the parse-miss branches already do this through metricOrUnavailable; the three early-return/exception branches did not. Record an unavailable sample for each runtime metric on those paths too, with the same broker labels, so the reconciler's unavailable-coverage guard keeps the active state instead of resolving it. Report URL: https://github.com/apache/rocketmq-dashboard/actions/runs/37095520338 With regards, GitHub Actions via GitBox
