[FLINK-30558] The metric 'numRestarts' reported in SchedulerBase will be overridden by metric 'fullRestarts'

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.16.0, 1.17.0
• Fix Version/s: 1.17.0, 1.16.1
• Component/s: Runtime / Metrics
• Labels: pull-request-available

Language

Java

• Assignee: xingbe
• Reporter: xingbe
• Created: 04/Jan/23 10:14
• Updated: 05/Jan/23 06:59
• Resolved: 05/Jan/23 06:59

Description

The method SchedulerBase#registerJobMetrics register metrics 'numRestarts' and 'fullRestarts' with the same metric object, as discussed in FLINK-30246, that will result in the loss of the metric 'numRestarts'.

metrics.gauge(MetricNames.NUM_RESTARTS, numberOfRestarts); 
metrics.gauge(MetricNames.FULL_RESTARTS, numberOfRestarts);

I have verified this problem via rest api /jobs/:jobid/metrics, and the response shows below, we can find that the metric 'numRestarts' is missing.

[{"id":"numberOfFailedCheckpoints"},{"id":"cancellingTime"},{"id":"lastCheckpointSize"},{"id":"totalNumberOfCheckpoints"},{"id":"lastCheckpointExternalPath"},{"id":"lastCheckpointRestoreTimestamp"},{"id":"failingTime"},{"id":"runningTime"},{"id":"uptime"},{"id":"restartingTime"},{"id":"initializingTime"},{"id":"numberOfInProgressCheckpoints"},{"id":"downtime"},{"id":"lastCheckpointProcessedData"},{"id":"numberOfCompletedCheckpoints"},{"id":"deployingTime"},{"id":"lastCheckpointFullSize"},{"id":"fullRestarts"},{"id":"createdTime"},{"id":"lastCheckpointDuration"},{"id":"lastCheckpointPersistedData"}]

Issue Links

• links to: GitHub Pull Request #21598