[
https://issues.apache.org/jira/browse/FLINK-21403?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Konstantin Knauf updated FLINK-21403:
-------------------------------------
Description:
The TableSinkITCase contains tests to ensure that a job submission fails if the
parallelism is exceedingly large.
The adaptive scheduler is having none of that and just downscales the job to
make it runnable, causing the tests to fail.
It may make sense to add a switch to disable downscaling (until we have a
proper range of accepted parallelism), as this issue will also affect e2e tests
where the start of processes takes so long that the resource timeout fires and
the job is run with a lower parallelism, which the tests don't handle well.
was:
The TableSinkITCase contains tests to ensure that a job submission fails if the
parallelism is exceedingly large.
The declarative scheduler is having none of that and just downscales the job to
make it runnable, causing the tests to fail.
It may make sense to add a switch to disable downscaling (until we have a
proper range of accepted parallelism), as this issue will also affect e2e tests
where the start of processes takes so long that the resource timeout fires and
the job is run with a lower parallelism, which the tests don't handle well.
> Some tests expect to fail if parallelism cannot be met
> ------------------------------------------------------
>
> Key: FLINK-21403
> URL: https://issues.apache.org/jira/browse/FLINK-21403
> Project: Flink
> Issue Type: Sub-task
> Components: Tests
> Affects Versions: 1.13.0
> Reporter: Chesnay Schepler
> Priority: Major
>
> The TableSinkITCase contains tests to ensure that a job submission fails if
> the parallelism is exceedingly large.
> The adaptive scheduler is having none of that and just downscales the job to
> make it runnable, causing the tests to fail.
> It may make sense to add a switch to disable downscaling (until we have a
> proper range of accepted parallelism), as this issue will also affect e2e
> tests where the start of processes takes so long that the resource timeout
> fires and the job is run with a lower parallelism, which the tests don't
> handle well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
