[FLINK-30870] Performance regressions notifications in Slack are cut off

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Component/s: Benchmarks
• Labels: pull-request-available
• Assignee: Yanfei Lei
• Reporter: Martijn Visser
• Created: 01/Feb/23 13:15
• Updated: 10/Feb/23 07:41
• Resolved: 10/Feb/23 07:41

Description

Example from today at https://apache-flink.slack.com/archives/C0471S0DFJ9/p1675253720571659

Performance regression
mapRebalanceMapSink.F27_UNBOUNDED median=17231.7398765 recent_median=16165.0549395
multiInputOneIdleMapSink median=11254.5329375 recent_median=10727.7280915
calculateRegionToRestart.BATCH median=12.881527 recent_median=12.096391
partitionRelease.BATCH median=23.2130145 recent_median=21.4858475
checkpointMultiInput median=2.6094395 recent_median=2.477736
checkpointSingleInput.UNALIGNED median=339.229515 recent_median=67.2695295
checkpointSingleInput.UNALIGNED_1 median=215.2789775 recent_median=40.1294965
fireProcessingTimers median=50.9977185 recent_median=44.0925955
globalWindow median=5459.689767 recent_median=5045.436655
<http…

As you can see, the last part is cut off with <htt... listed. I'm not sure if this is because there are more regressions, but they aren't posted to Slack (could be) or if this is just a mistake in the output and the globalWindow was the last regression to be reported. It would be great if this could be validated.

Issue Links

• links to: GitHub Pull Request #65