[
https://issues.apache.org/jira/browse/CASSANDRA-14566?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17390544#comment-17390544
]
Benjamin Lerer commented on CASSANDRA-14566:
--------------------------------------------
[~pougetat] Sorry for the delay.
I had a look to the patch and I have the following comments:
# The {{onlyPurgeRepTomb}} variable visibily across threads is not guarantee as
{{setStrategy}} calls are guarded by the {{writeLock}} but calls to
{{onlyPurgeRepairedTombstones}} are not guarded by any lock and the variable is
not volatile.
# I am not super familiar with the compaction part of the code base but I
believe that we should guarante that the {{params}} and the cached value should
be set together to guaranty thread safety.
I will put the ticket back in {{open}} state. Do not hesitate to put it back in
{{Patch Available}} once the comments have been addressed.
> Cache CSM.onlyPurgeRepairedTombstones()
> ---------------------------------------
>
> Key: CASSANDRA-14566
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14566
> Project: Cassandra
> Issue Type: Improvement
> Components: Local/Compaction
> Reporter: Stefan Podkowinski
> Assignee: Thomas Pouget-Abadie
> Priority: Low
> Labels: lhf, performance
> Attachments: 14566-3.11.txt
>
>
> We currently call {{CompactionStrategyManager.onlyPurgeRepairedTombstones()}}
> *a lot* during compaction, I think at least for every key. So we should
> probably cache the value, instead of constantly reading from a volatile and
> calling parseBoolean.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
