You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Semyon Danilov (Jira)" <ji...@apache.org> on 2023/01/25 12:44:00 UTC
[jira] [Commented] (IGNITE-18446) Fix busylock usage in ClusterManagementGroupManager
[ https://issues.apache.org/jira/browse/IGNITE-18446?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17680602#comment-17680602 ]
Semyon Danilov commented on IGNITE-18446:
-----------------------------------------
I discovered that CMG doesn't block node stopping, but meta storage does. Fixed it for meta storage and created a new task for other components https://issues.apache.org/jira/browse/IGNITE-18636
> Fix busylock usage in ClusterManagementGroupManager
> ---------------------------------------------------
>
> Key: IGNITE-18446
> URL: https://issues.apache.org/jira/browse/IGNITE-18446
> Project: Ignite
> Issue Type: Bug
> Reporter: Andrey Mashenkov
> Assignee: Semyon Danilov
> Priority: Major
> Labels: ignite-3
> Time Spent: 10m
> Remaining Estimate: 0h
>
> As for now we chain operation future to a joinFuture under busy lock, but operation futures completion is not wrapped with busy lock.
> Startpoint is ClusterManagementGroup.metaStorageNodes(): in readClusterState() method we run a raft command, but raft service might not be aware of node is stopping (as well as lower level components).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)