[
https://issues.apache.org/jira/browse/CASSANDRA-5248?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13577633#comment-13577633
]
Yuki Morishita commented on CASSANDRA-5248:
-------------------------------------------
Wait, we do want to delete even if key exits in other sstable but the sstable
timestamp was older than max deletion time, don't we?
The patch changes the behavior to not delete even if such key exists.
> Fix timestamp-based tomstone removal logic
> ------------------------------------------
>
> Key: CASSANDRA-5248
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5248
> Project: Cassandra
> Issue Type: Bug
> Affects Versions: 1.2.1
> Reporter: Sylvain Lebresne
> Assignee: Sylvain Lebresne
> Fix For: 1.2.2
>
> Attachments: 5248.txt
>
>
> Quoting the description of CASSANDRA-4671:
> {quote}
> In other words, we should force CompactionController.shouldPurge() to return
> true if min_timestamp(non-compacted-overlapping-sstables) >
> max_timestamp(compacted-sstables)
> {quote}
> but somehow this was translating in the code to:
> {noformat}
> if (sstable.getBloomFilter().isPresent(key.key) && sstable.getMinTimestamp()
> >= maxDeletionTimestamp)
> return false;
> {noformat}
> which, well, is reversed.
> Attaching the trivial patch to fix. I note that we already had a test that
> catched this (CompactionsTest.testDontPurgeAccidentaly), but that test was
> racy in that most of the time the compaction was done in the same second than
> the removal done prior to that and thus the compaction wasn't considering the
> tombstone gcable even though gcgrace was 0. I've already pushed the addition
> of a 1 second delay to make sure the patch reliably catch this bug.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
