[
https://issues.apache.org/jira/browse/IGNITE-9737?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16632210#comment-16632210
]
Andrey Kuznetsov commented on IGNITE-9737:
------------------------------------------
Critical threads monitoring should be configured in a more flexible way than it
was originally suggested in issue description.
# Separate timeout setting should be introduced for {{SYSTEM_WORKER_BLOCKED}}
detection in {{IgniteConfiguration}}, falling back to
{{IgniteConfiguration.failureDetectionTimeout}} if omitted. This can be
overridden by a system property or changed in runtime by newly created
management bean. Timeout value of 0 should denote infinite timeout, thus it has
an effect of detection disabling.
# Separate timeout should be introduced for checkpoint read lock acquisition in
the same manner.
# When timed out, checkpoint read lock acquisition should not throw an
exception unless failure handler invalidated the node. This will guarantee
neutral behavior of {{NoOpFailureHandler}}.
> Ignite WatchDog service should be configurable
> ----------------------------------------------
>
> Key: IGNITE-9737
> URL: https://issues.apache.org/jira/browse/IGNITE-9737
> Project: Ignite
> Issue Type: Bug
> Reporter: Nikolay Izhikov
> Assignee: Andrey Kuznetsov
> Priority: Blocker
> Fix For: 2.7
>
>
> At the moment, there is no way to disable Ignite WatchDog service from config
> or JVM option.
> In any corner case or bug in that feature Ignite can become fully unusable
> due to unpredictable shutdown.
> We should provide a way to enable/disable this feature from config or from
> JVM option.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
