Sample not requeued or ack'ed on sample publishing failure

Bug #1486660 reported by Rohit Jaiswal
18
This bug affects 3 people
Affects Status Importance Assigned to Milestone
Ceilometer
Triaged
Medium
Rohit Jaiswal

Bug Description

When the sample pipeline sink has a single publisher configured and persistence or publishing fails, the configured sample publisher will propagate the Exception to SampleSink in pipeline:https://github.com/openstack/ceilometer/blob/master/ceilometer/pipeline.py#L485, where its caught and logged/handled, but not raised again.

The pipeline code assumes multiple publishers as the default case, when in a single publisher case, an error from publisher means the sample is likely lost and should be requeued.

Notification agent should support ack_on_sample_error similar to supporting ack_on_event_error and allow sample requeuing.

Changed in ceilometer:
assignee: nobody → Rohit Jaiswal (rohit-jaiswal-3)
gordon chung (chungg)
Changed in ceilometer:
status: New → Triaged
importance: Undecided → Medium
Revision history for this message
Rohit Jaiswal (rohit-jaiswal-3) wrote :

The fix for this will result in a requeue feature for samples in the notification agent which should be disabled by default and should only be enabled when pipeline is configured with single publisher for samples. It should also handle the case where samples are published to pipeline queues in HA scenario. Samples from api and polling source should be requeued on the sample queue.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Duplicates of this bug

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.