[
https://issues.apache.org/jira/browse/CASSANDRA-18222?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Capwell updated CASSANDRA-18222:
--------------------------------------
Description:
The accord integration tests are currently flaky due to Preempt happening (we
convert that to a timeout). We need to address this to make sure the tests are
stable.
One solution may be make SimpleProgressLog’s frequency run different for
jvm-dtest (likely to depend on CASSANDRA-18221 ), this would be similar to what
we already do as timeouts maybe larger in jvm-dtest due to the extra complexity
running multiple processes in the same JVM. Another solution may be to have
the client coordinator “attach” to the new coordinator, this would avoid the
user being impacted by this preempt when the coordinator is healthy.
Example:
https://app.circleci.com/pipelines/github/dcapwell/cassandra/1822/workflows/0d92898f-ad7d-4a7d-9198-3a7ce844ee93/jobs/16464
was:
The accord integration tests are currently flaky due to Preempt happening (we
convert that to a timeout). We need to address this to make sure the tests are
stable.
One solution may be make SimpleProgressLog’s frequency run different for
jvm-dtest (likely to depend on CASSANDRA-18221 ), this would be similar to what
we already do as timeouts maybe larger in jvm-dtest due to the extra complexity
running multiple processes in the same JVM. Another solution may be to have
the client coordinator “attach” to the new coordinator, this would avoid the
user being impacted by this preempt when the coordinator is healthy.
> Fix flaky test
> org.apache.cassandra.distributed.test.accord.AccordIntegrationTest
> ---------------------------------------------------------------------------------
>
> Key: CASSANDRA-18222
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18222
> Project: Cassandra
> Issue Type: Bug
> Components: Accord, CI
> Reporter: David Capwell
> Priority: Normal
> Fix For: 4.x
>
>
> The accord integration tests are currently flaky due to Preempt happening (we
> convert that to a timeout). We need to address this to make sure the tests
> are stable.
> One solution may be make SimpleProgressLog’s frequency run different for
> jvm-dtest (likely to depend on CASSANDRA-18221 ), this would be similar to
> what we already do as timeouts maybe larger in jvm-dtest due to the extra
> complexity running multiple processes in the same JVM. Another solution may
> be to have the client coordinator “attach” to the new coordinator, this would
> avoid the user being impacted by this preempt when the coordinator is
> healthy.
> Example:
> https://app.circleci.com/pipelines/github/dcapwell/cassandra/1822/workflows/0d92898f-ad7d-4a7d-9198-3a7ce844ee93/jobs/16464
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
