[
https://issues.apache.org/jira/browse/CASSANDRA-11834?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Ellis updated CASSANDRA-11834:
---------------------------------------
Resolution: Fixed
Fix Version/s: 2.2.7
Status: Resolved (was: Patch Available)
committed to 2.1 and 2.2
> Don't compute expensive MaxPurgeableTimestamp until we've verified there's an
> expired tombstone
> -----------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-11834
> URL: https://issues.apache.org/jira/browse/CASSANDRA-11834
> Project: Cassandra
> Issue Type: Bug
> Components: Compaction
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 2.1.15, 2.2.7
>
> Attachments: 11834.txt
>
>
> In LCR's get reduced, we currently do this:
> {code}
> if (t.timestamp() < getMaxPurgeableTimestamp() &&
> t.data.isGcAble(controller.gcBefore))
> {code}
> Should call the expensive getMaxPurgeableTimestamp only after we have called
> the cheap isGcAble.
> Marking this as a bug since it can cause pathological performance problems
> (CASSANDRA-11831).
> Have verified that this is not a problem in 3.0 (CompactionIterator does the
> check in the correct order).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
