[ 
https://issues.apache.org/jira/browse/CASSANDRA-10836?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15115044#comment-15115044
 ] 

Sylvain Lebresne commented on CASSANDRA-10836:
----------------------------------------------

I strongly suspect this is an issue with the tests. All those tests are doing 
conditional updates but use non SERIAL reads, and since the consistency for 
both commit and read defaults to ONE, it is expected that there is no specific 
consistency guarantees and the flapping is not surprising. I've created a [pull 
request|https://github.com/riptano/cassandra-dtest/pull/770] to fix that and 
use SERIAL reads. There was also a mix of conditional and non-conditional 
updates in those tests, which is equally dodgy, so the pull request changes 
those so we can hopefully get more reliable testing.

I've checked that all the modified test pass with the change, but given the 
tests were flapping somewhat irregularly, I think we'll have to commit and wait 
to see if this does fix the issue on the long run (but I'm confident we _at 
least_ need the changes of the pull request if we want reliable test).

> Data integrity flappers in upgrade_tests dtests
> -----------------------------------------------
>
>                 Key: CASSANDRA-10836
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-10836
>             Project: Cassandra
>          Issue Type: Sub-task
>            Reporter: Jim Witschey
>            Priority: Critical
>             Fix For: 3.0.x, 3.x
>
>
> A number of the {{upgrade_tests}} in dtest are flapping with data integrity 
> bugs. For instance:
> http://cassci.datastax.com/view/cassandra-3.0/job/cassandra-3.0_dtest/422/testReport/upgrade_tests.cql_tests/TestCQLNodes3RF3/conditional_delete_test/history/
> http://cassci.datastax.com/view/cassandra-3.0/job/cassandra-3.0_dtest/421/testReport/upgrade_tests.cql_tests/TestCQLNodes3RF3/cas_and_list_index_test/history/
> This needs further digging to get good debug information; it could well be, 
> e.g., a race condition in the test. I have not reproduced locally.
> /cc [~philipthompson]



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to