[
https://issues.apache.org/jira/browse/ZOOKEEPER-4927?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17950165#comment-17950165
]
Kezhu Wang commented on ZOOKEEPER-4927:
---------------------------------------
I found two mails about backport:
# [https://lists.apache.org/thread/q1wfmc1bomswjvolmkwv9bfxpqgmhlxg]
# [https://lists.apache.org/thread/tdoc8wco2606qcb4myl5fttnhwl9dmwf]
Currently, 3.9 is the "current" release so it can accept improvement
commits(a.k.a. not bug fix). I think ZOOKEEPER-4829 is small and could be
considered as an improvement, but [the
commit|https://github.com/apache/zookeeper/pull/2178/files#diff-c184aadd789e20edde3f3f4a95c8d6f2ae0802c90fae2ed11df8878062fd0a1fL872]
itself renamed a public class
[QuorumPeerConfig|https://github.com/apache/zookeeper/blob/ec0aa27b1053c839333038efbada98b6a3504728/zookeeper-server/src/main/java/org/apache/zookeeper/server/quorum/QuorumPeerConfig.java#L63].
I think this is not allowed according to our [Backward
Compatibility|https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=25200595#ReleaseManagement-BackwardCompatibility].
So we may have to fix that before backport.
> Support more fine tunable purgeInterval
> ---------------------------------------
>
> Key: ZOOKEEPER-4927
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-4927
> Project: ZooKeeper
> Issue Type: Improvement
> Components: server
> Affects Versions: 3.9.3
> Reporter: Li Wang
> Priority: Major
>
> A zookeeper instance went down in prod because it ran out of disk space. It
> turned out that the purge task was not able to keep up with the rate of
> snapshot taken. A new snapshot was taken every a couple of mins.
> Too snapshots were generate during autoPurge.purgeInterval. Since the unit is
> hour, so the min internal is 1 hour.
> To support writes heavy use case, we would need to support more fine tuned
> purge interval. For example, in minutes.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
