[
https://issues.apache.org/jira/browse/FLINK-25869?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17500017#comment-17500017
]
Piotr Nowojski commented on FLINK-25869:
----------------------------------------
After further offline discussions we reached a consensus that at the moment we
do not want to create such stress tests.
1. It seems unfeasible to add test coverage for FLINK-25728. Some estimations
suggest that for a reliable test coverage we would need to run a single test
for this bug continuously for many hours.
2. Another potential candidate {{UnalignedCheckpointStressITCase}}, doesn't
make a difference from coverage perspective if it will be executed for 10
minutes once a day, or 20 seconds 30 times a day (per each master commit/PR),
while it's better to have as many tests as possible per commit
3. Another idea of implementing some memory usage regression benchmarks, like
running some job for 5 minutes and reporting average used memory, seemed more
realistic, but that wouldn't help to cover FLINK-25728. At the same time we do
not see demand/pressure from Flink users to actually focus our efforts on
memory consumption.
> Create a stress test cron build
> -------------------------------
>
> Key: FLINK-25869
> URL: https://issues.apache.org/jira/browse/FLINK-25869
> Project: Flink
> Issue Type: Improvement
> Components: Test Infrastructure, Tests
> Affects Versions: 1.15.0
> Reporter: Piotr Nowojski
> Priority: Not a Priority
>
> We propose creating some kind of stress test cron job dedicated for
> periodically running longer lasting stress tests. For example trying to
> expose memory leaks.
> First idea to test could be to provide test coverage for FLINK-25728.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
