[jira] [Commented] (CASSANDRA-7414) After cleanup we can end up with non-compacting high level sstables

["Yuki Morishita (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-7414?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14036937#comment-14036937 ] 

Yuki Morishita commented on CASSANDRA-7414:
-------------------------------------------

Hmm, the problem here is not clear to me. Is it that it will be never compacted because compaction score changes when cleanup/sstable size change? I think if we continue to flush SSTables, eventually levels are filled up and reach compaction threshold.

Some comments to the patch:

* getOverlappingStarvedSSTables: you can pull calculating min/max boundary out of the loop.
* getOverlappingStarvedSSTables: why don't just add sstable to withStarvedCandidate instead of union set of 1 sstable.

> After cleanup we can end up with non-compacting high level sstables
> -------------------------------------------------------------------
>
>                 Key: CASSANDRA-7414
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7414
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Marcus Eriksson
>            Assignee: Marcus Eriksson
>            Priority: Minor
>         Attachments: 0001-Include-high-level-sstables-in-lower-level-compactio.patch
>
>
> If we run cleanup (or increase sstable size) on a node with LCS, we could end up with a bunch of sstables in higher levels that are "never" compacted.



--
This message was sent by Atlassian JIRA
(v6.2#6252)