[
https://issues.apache.org/jira/browse/TUBEMQ-253?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17143825#comment-17143825
]
Guocheng Zhang commented on TUBEMQ-253:
---------------------------------------
This problem should be introduced when the TUBEMQ-149 problem was modified:
after the indexPartition object was changed from BlockingQueue to
ConcurrentLinkedQueue, in the pushSelect method, after the previous take method
was not provided, the interface was changed to the poll mode, which caused the
problem. Some businesses use the pull mode, so the problem is not discovered in
time:
!screenshot-1.png!
> tube-consumer fetch-worker cpu used too high
> --------------------------------------------
>
> Key: TUBEMQ-253
> URL: https://issues.apache.org/jira/browse/TUBEMQ-253
> Project: Apache TubeMQ
> Issue Type: Bug
> Components: Client
> Reporter: yupili
> Assignee: Guocheng Zhang
> Priority: Major
> Labels: pull-request-available
> Attachments: screenshot-1.png
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> java tube-client sdk 0.3.0
> start 4 fetch-worker which are all runnable status and cause cpu occupy 390+!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
