[
https://issues.apache.org/jira/browse/IGNITE-22398?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17854930#comment-17854930
]
Roman Puchkovskiy commented on IGNITE-22398:
--------------------------------------------
CriticalWorkerWatchdog's purpose is to detect situations when critical threads
are stalled and this poses the cluster stability at danger (the danger so
imminent that we prefer to crash the node). But TrackableNetworkMessageHandler
is about milder conditions: detecting message handling that does not fit in
5ms; it's a warning, not an error like with CriticalWorkerWatchdog.
Also, the watchdog is not designed to work with delays like 5ms.
It seems that it makes sense to keep the mechanisms separate.
> Rework network threads' lags detection using the critical workers watchdog
> --------------------------------------------------------------------------
>
> Key: IGNITE-22398
> URL: https://issues.apache.org/jira/browse/IGNITE-22398
> Project: Ignite
> Issue Type: Improvement
> Reporter: Denis Chudov
> Priority: Major
> Labels: ignite-3
>
> *Motivation*
> CriticalWorkerWatchdog should become the common approach to detect the
> critical threads' lags.
> *Definition of done*
> * Network threads' lags and slow processing of messages should be reworked
> using CriticalWorkerWatchdog;
> * TrackableNetworkMessageHandler should be removed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
