You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ratis.apache.org by "Lokesh Jain (Jira)" <ji...@apache.org> on 2019/11/25 09:22:00 UTC
[jira] [Commented] (RATIS-727) Garbage collection due to same
request retries on a follower
[ https://issues.apache.org/jira/browse/RATIS-727?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16981404#comment-16981404 ]
Lokesh Jain commented on RATIS-727:
-----------------------------------
In the cluster when heap dump was taken one of the followers was far behind other members of the group. Therefore we were seeing a lot of NotReplicatedExceptions. Currently when a raft client sees a NotReplicatedException it changes the leader and resends the existing requests including write chunk requests. This is currently tracked in RATIS-709 and might be a reason for garbage generation in the follower.
> Garbage collection due to same request retries on a follower
> ------------------------------------------------------------
>
> Key: RATIS-727
> URL: https://issues.apache.org/jira/browse/RATIS-727
> Project: Ratis
> Issue Type: Bug
> Components: client
> Reporter: Lokesh Jain
> Assignee: Hanisha Koneru
> Priority: Blocker
>
> In a heap dump it could be seen that a client request retries on the same follower multiple times and every time the request is rejected with a NotLeaderException. In case of Ozone it is a WriteChunk request which leads to garbage collection of 16MB for every request. In the heap dump a client request retries multiple times leading to garbage collection of ~100MB.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)