[
https://issues.apache.org/jira/browse/CASSANDRA-17217?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17509087#comment-17509087
]
Yifan Cai commented on CASSANDRA-17217:
---------------------------------------
The flaky test is likely cause by the prior read query fails earlier than the
completion of the read command execution in each replica. A read is failed as
soon as one of the read command fails, and a failure is returned to client.
However, there could be still inflight read commands in the other replicas.
They will eventually fail and increment the counter.
For the test in question, once the size tracking is turned off, new queries
should not update the metrics, but the flakiness is cause by the prior read.
There is nothing wrong with the behavior. Probably either wait for some time
between each read query to make sure all replicas complete their read command.
Or even more clean, test without changing the size tracking in between. It
basically breaks the existing test into 2, read with size tracking and w/o.
> LocalReadSizeWarningTest#failThresholdSinglePartition is flaky
> --------------------------------------------------------------
>
> Key: CASSANDRA-17217
> URL: https://issues.apache.org/jira/browse/CASSANDRA-17217
> Project: Cassandra
> Issue Type: Bug
> Components: Test/unit
> Reporter: Andres de la Peña
> Assignee: Bernardo Botella Corbi
> Priority: Normal
> Fix For: 4.x
>
>
> The test
> {{org.apache.cassandra.distributed.test.trackwarnings.LocalReadSizeWarningTest#failThresholdSinglePartition}}
> seems to be flaky.
> It was discovered during CASSANDRA-17195 with [this CI
> run|https://app.circleci.com/pipelines/github/adelapena/cassandra/1217/workflows/b366e352-0862-485a-acdc-5b75fe1ef575/jobs/11188].
> The failure can be reproduced running the test repeatedly for trunk, as it
> can be seen
> [here|https://app.circleci.com/pipelines/github/adelapena/cassandra/1219/workflows/a0c1fb7f-8a07-4fcf-97a4-7caf207dad78].
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
