[jira] [Updated] (CASSANDRA-14244) Some tests in read_repair_test are flakey

["Sam Tunnicliffe (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CASSANDRA-14244?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sam Tunnicliffe updated CASSANDRA-14244:
----------------------------------------
    Status: Patch Available  (was: Open)

DTest PR for this here: https://github.com/apache/cassandra-dtest/pull/20
One additional change I made, aside from setting the table options, was to disable dynamic snitch as this could also cause a coordinator to direct a read command to another replicwhen it could execute it itself.

CI is running against that dtest branch and I haven't seen a failure in over 50 runs locally
https://circleci.com/workflow-run/69c20512-3407-45e9-887a-5a7b2f7ea95b

> Some tests in read_repair_test are flakey
> -----------------------------------------
>
>                 Key: CASSANDRA-14244
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-14244
>             Project: Cassandra
>          Issue Type: Test
>          Components: Testing
>            Reporter: Sam Tunnicliffe
>            Assignee: Sam Tunnicliffe
>            Priority: Major
>              Labels: dtest
>
> Since being refactored for CASSANDRA-14134, {{test_alter_rf_and_run_read_repair}} and {{test_read_repair_chance}} in {{read_repair_test.TestReadRepair}} are flakey and regularly fail on all branches. The problem is that the inital setup of these two tests doesn't explicitly set the {{read_repair_chance}} or {{dclocal_read_repair_chance}} properties on the test table. As a consequence, read repairs are sometimes probabilistically triggered and query results don't match the expectations.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org