[jira] [Commented] (IGNITE-22397) Improve the CriticalWorkerWatchdog to make it able to measure different delays for different threads

["Roman Puchkovskiy (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-22397?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17854932#comment-17854932 ] 

Roman Puchkovskiy commented on IGNITE-22397:
--------------------------------------------

In general, it might make sense to configure the allowed delay per worker, but TrackableNetworkMessageHandler mentioned here seems to need a different mechanism, see my comment https://issues.apache.org/jira/browse/IGNITE-22398?focusedCommentId=17854930&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17854930

> Improve the CriticalWorkerWatchdog to make it able to measure different delays for different threads
> ----------------------------------------------------------------------------------------------------
>
>                 Key: IGNITE-22397
>                 URL: https://issues.apache.org/jira/browse/IGNITE-22397
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Denis Chudov
>            Priority: Major
>              Labels: ignite-3
>
> *Motivation*
> For now, the critical worker is some thread that cannot be blocked for more than X ms, and the value X is the same for every worker and configured with 
> CriticalWorkersConfiguration#maxAllowedLag. We have different mechanisms for other threads like network ones (see 
> TrackableNetworkMessageHandler). Using the CriticalWorkerWatchdog we can introduce the common approach for every case.
>  
> *Definition of done*
>  - we are able to configure maxAllowedLag per thread (per pool);
>  - we are able to detect small lags of critical threads (such as 5 ms which is used in TrackableNetworkMessageHandler) using CriticalWorkerWatchdog.



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