[
https://issues.apache.org/jira/browse/CASSANDRA-18211?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17682970#comment-17682970
]
Jacek Lewandowski commented on CASSANDRA-18211:
-----------------------------------------------
[~smiklosovic] I've also created a small PR with my findings -
https://github.com/apache/cassandra/pull/2124/files
It also makes SnapshostsTest much faster, 3x times on my local machine. There
are two main changes which affects performance of the test:
- it uses gossip, which I think is completely unrelated to what that test aims
to verify
- the method which verifies snapshot existence does it every 1 second, and what
is more important, start verifying after 1 second, so it lasts at least 1
second. In after test hook we remove all the snapshots one by one and we verify
that using that method - since we have 12 snapshots to verify, it adds at least
12 seconds to each test case
> Possible race between automatic and manual snapshot cleanup
> -----------------------------------------------------------
>
> Key: CASSANDRA-18211
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18211
> Project: Cassandra
> Issue Type: Bug
> Components: Local/Snapshots
> Reporter: Jacek Lewandowski
> Assignee: Jacek Lewandowski
> Priority: Normal
> Fix For: 4.1.x, 4.x
>
>
> Automatic deletion of expired snapshots maintained by {{SnapshotManager}} can
> race with manual snapshot removal in a way an exception like
> {{NoSuchFileException}} is thrown.
> It is because the snapshot directory existence is checked and deleted if it
> exists as a non-atomic operation. Since we can potentially have two threads
> attempting to do that at the same time (automatic and manual snapshot
> removal) it may lead to a race in rare situations.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
