[
https://issues.apache.org/jira/browse/CASSANDRA-8592?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14352716#comment-14352716
]
Stefania commented on CASSANDRA-8592:
-------------------------------------
Tyler, I started working on a test:
https://github.com/stef1927/cassandra-dtest/tree/8592.
Other than creating a mutation bigger than half of
{{commitlog_segment_size_in_mb}} I could not find any other way to fail a
mutation without manually adding a fake exception. I spent some time trying to
remove the permissions to the commit log, which does work, but unfortunately in
this case we wait for a segment to become available and we time out, we do no
throw anything. Any other ideas how to fail mutations? The drawback with using
{{commitlog_segment_size_in_mb}} is that the mutation fails everywhere, locally
or on replicas, it would be good to have only a set of replicas failing but
it's otherwise adequate to test the code path that was modified.
Also, do we need to test with a V4 client? The dtest is still getting a
{{WriteTimeout}} because of the conversion to legacy exceptions which happens
if the protocol version is less than or equal to 3.
The code is available here and it is pretty much complete as far as I
understand the problem so I am looking for comments:
https://github.com/stef1927/cassandra/tree/8592. Are you happy to be the
reviewer or shall I find someone else?
> Add WriteFailureException
> -------------------------
>
> Key: CASSANDRA-8592
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8592
> Project: Cassandra
> Issue Type: New Feature
> Components: API, Core
> Reporter: Tyler Hobbs
> Assignee: Stefania
> Labels: client-impacting
> Fix For: 3.0
>
>
> Similar to what CASSANDRA-7886 did for reads, we should add a
> WriteFailureException and have replicas signal a failure while handling a
> write to the coordinator.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
