[
https://issues.apache.org/jira/browse/CASSANDRA-20509?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17942003#comment-17942003
]
Ariel Weisberg commented on CASSANDRA-20509:
--------------------------------------------
The test itself is not particularly high value in terms of its implementation.
It's testing for a very specific NIO problem where NIO has a thread local
buffer cache and working with large columns can cause NIO to allocate a large
DBB per thread which is basically the high watermark for any message sent. NIO
allocates this buffer if you try to read from a socket/socketchannel with a
heap bytebuffer.
Does Netty still use NIO or its own JNI based thing? I haven't look at Netty in
a long time. It's also possible NIO changed it's memory management strategy
when using heap byte buffers.
It's a regression test so the answer to do we want it is how badly do we want
to avoid re-introducing the issue of working with large columns causing a lot
of direct memory usage?
> Test failure: largecolumn_test.TestLargeColumn.test_cleanup
> -----------------------------------------------------------
>
> Key: CASSANDRA-20509
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20509
> Project: Apache Cassandra
> Issue Type: Bug
> Components: Test/dtest/python
> Reporter: Dmitry Konstantinov
> Assignee: Stefan Miklosovic
> Priority: Normal
>
> https://ci-cassandra.apache.org/job/Cassandra-trunk/2086/#showFailuresLink
> TestLargeColumn.test_cleanup starts to fail since 2086 build.
> It was flaky some time ago but I fixed it and it was stable: CASSANDRA-19702
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
