[
https://issues.apache.org/jira/browse/IGNITE-17945?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17647061#comment-17647061
]
Maksim Timonin commented on IGNITE-17945:
-----------------------------------------
[~zstan] hi! Thanks for participating in this ticket!
I think such fix can be implemented without discussion on the devlist. Because
the place, reason, message, log level, param dimension (milliseconds) don't
change. Then I suppose no public API changed. We're only going to add a minor
clarification.
But I agree with your comment about text message. I actually missed that
recursion. Is it better to change the message to
"Possible too long JVM pause. Possible reasons: GC *or other JVM internal*
{*}pauses{*}, hardware temporary freeze, bulk loading".
This message should notify user that GC is not an only reason for this pause.
WDYT?
> Change message in case watchdog thread observes freeze
> ------------------------------------------------------
>
> Key: IGNITE-17945
> URL: https://issues.apache.org/jira/browse/IGNITE-17945
> Project: Ignite
> Issue Type: Improvement
> Reporter: Dmitry Pavlov
> Assignee: Julia Bakulina
> Priority: Minor
> Labels: ise, newbie
> Fix For: 2.15
>
> Time Spent: 40m
> Remaining Estimate: 0h
>
> In Long JVM pause detector we warn user that the thread wasn't alive for a
> threshold
> Possible too long JVM pause: 500+ milliseconds.
> But it is often treated as a GC pause.
> We can change that warning to
> Possible too long JVM pause: 500+ milliseconds. (GC pause or hardware
> temporary freeze)
> To enlist possible reasons behind that behaviour
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
