[
https://issues.apache.org/jira/browse/IGNITE-24501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17929083#comment-17929083
]
Philipp Shergalis commented on IGNITE-24501:
--------------------------------------------
LGTM
> Introduce a separate disruptor pool for CMG
> -------------------------------------------
>
> Key: IGNITE-24501
> URL: https://issues.apache.org/jira/browse/IGNITE-24501
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Time Spent: 20m
> Remaining Estimate: 0h
>
> CMG shares disruptor pools with partitions. As a result, during startup, when
> partitions initiate recovery, they create a substantial load on JRaft
> disruptors. This might lead a node not being able to handle CMG commands (for
> example, to JoinReady command not being able to be executed in time).
> A similar problem was earlier solved by using a separate disruptor pool for
> Metastorage, take a look at MetaStorageManagerImpl#initializeMetastorage().
> Here is what needs to be done:
> # Use a separate disruptor for CMG
> # When submitting a JoinReadyRequest to the CMG, wait for response forever
> (and not the default 10 seconds) - see CmgRaftService#completeJoinCluster()
> [raftService.run(command, RaftCommandRunner.NO_TIMEOUT) should be used]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
