Skip to content

Fix status message reporting

Afonso Mukai requested to merge DM-1383_fix_status_message_reporting_dies into master

Created by: matthew-d-jones

Issue

DM-1383

The bug is that status message reporting stops after several hours of File Writer runtime. This was observed at V20.

The bug was caused by insufficient calls to RdKafka::Producer.poll(). This results in successful publish events not getting handled, therefore RdKafka does not clear the messages from the producer queue, it eventually fills up completely and at that point message publishing fails.

Description of work

I've added a test to Producer_tests.cpp which checks that poll is called each time we call produce. I've added a non-blocking call to poll in produce. Note, there is no significant performance hit from this, it only polls the local producer, it does not do anything across the network.

Nominate for Group Code Review

  • Nominate for code review

Merge request reports