[
https://issues.apache.org/jira/browse/RATIS-1622?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17565433#comment-17565433
]
Riguz Lee commented on RATIS-1622:
----------------------------------
Hi [~szetszwo] ,
In GrpcLogAppender,
{code:java}
public long getWaitTimeMs() {
if (haveTooManyPendingRequests()) {
return getHeartbeatWaitTimeMs(); // Should wait for a short time
} else if (shouldSendAppendEntries()) {
return 0L;
}
return Math.min(10L, getHeartbeatWaitTimeMs());
} {code}
it's could return 0L if the follower's commit index is behind, however if the
follower is down, then it will keep looping, right? Here's a screenshot using
arthas, which might be a evidence of that, but it might also be incorrect since
I got different results each time.
!image-2022-07-12-19-20-06-014.png|width=438,height=191!
> High cpu usage of LogAppenderDaemon
> -----------------------------------
>
> Key: RATIS-1622
> URL: https://issues.apache.org/jira/browse/RATIS-1622
> Project: Ratis
> Issue Type: Improvement
> Components: server
> Reporter: Tsz-wo Sze
> Assignee: Riguz Lee
> Priority: Major
> Attachments: image-2022-07-12-15-39-40-948.png,
> image-2022-07-12-15-40-29-685.png, image-2022-07-12-15-50-40-036.png,
> image-2022-07-12-19-20-06-014.png
>
>
> As reported by Riguz Lee in
> https://lists.apache.org/thread/p5hq3nr1zqpkgd3fr0vyx9j5d2lx6gf9 ,
> LogAppenderDaemon has a high cpu usage.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
