[ 
https://issues.apache.org/jira/browse/KYLIN-5406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17683696#comment-17683696
 ] 

longfeiJiang commented on KYLIN-5406:
-------------------------------------

*RootCause*

The scheduled polling task execution (FetcherRunner) uses the thread pool 
scheduleWithFixedDelay polling execution. When polling is a task using this 
method, when an exception occurs in the polling thread, it will not continue to 
execute the next thread polling

!image-2023-02-03-14-32-43-107.png!

 

*Fix Design*

Catch the exception executed by the thread to avoid the exception thrown by the 
current execution, which will cause the next execution to fail to execute 
normally

 

Also add the exception information to log, which is convenient for 
troubleshooting the cause of the execution exception.

> longrunning - All build tasks pending during execution
> ------------------------------------------------------
>
>                 Key: KYLIN-5406
>                 URL: https://issues.apache.org/jira/browse/KYLIN-5406
>             Project: Kylin
>          Issue Type: Bug
>    Affects Versions: 5.0-alpha
>            Reporter: longfeiJiang
>            Assignee: longfeiJiang
>            Priority: Major
>             Fix For: 5.0-alpha
>
>         Attachments: image-2023-02-02-16-55-59-701.png, 
> image-2023-02-03-14-32-43-107.png
>
>
> During the longrunning of Kylin , the environment resources are normal, but 
> all build tasks are in the pending state
> h5. {color:#0052cc}A{color}ppearance
> 1. Kylin server is single-node
> 2. Kylin has 100 projects:
> only 30 project can be scheduled to execute tasks normally, 70 project cannot 
> be executed, tasks are in pending state
> 3. 70 project does not have a log in the Kylin log for the timed polling task 
> execution (FetcherRunner):
>     Success: "Job Status in project {}: {} should running, {} actual running, 
> {} stopped, {} ready, {} already succeed, {} error, {} discarded, {} 
> suicidal, {} others"
>     Failed: Job Fetcher caught a exception
>  
> h5. Summary
> 70 projects without FetcherRunner are polling properly, resulting in all 
> build tasks in pending state



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to