You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@kudu.apache.org by "Grant Henke (Jira)" <ji...@apache.org> on 2020/06/02 12:56:00 UTC
[jira] [Commented] (KUDU-1701) Reduce contention in
CatalogManager::ScopedLeaderSharedLock
[ https://issues.apache.org/jira/browse/KUDU-1701?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17123748#comment-17123748 ]
Grant Henke commented on KUDU-1701:
-----------------------------------
[~aserbin] Is this resolved by https://github.com/apache/kudu/commit/14912a1fd78ba7cf4d62bf934ae64d6f6f229ee6 ?
> Reduce contention in CatalogManager::ScopedLeaderSharedLock
> ------------------------------------------------------------
>
> Key: KUDU-1701
> URL: https://issues.apache.org/jira/browse/KUDU-1701
> Project: Kudu
> Issue Type: Improvement
> Components: master
> Affects Versions: 1.0.0
> Reporter: Todd Lipcon
> Assignee: Alexey Serbin
> Priority: Minor
>
> CatalogManager::ScopedLeaderSharedLock::ScopedLeaderSharedLock() currently holds a spinlock while accessing consensus->ConsensusState(). That call makes a copy of a protobuf, which requires allocation and a later deallocation. Every master lookup RPC needs to go through this path, so this can become a contention point under heavy multi-client load.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)