[
https://issues.apache.org/jira/browse/IGNITE-22180?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksandr Polovtsev updated IGNITE-22180:
-----------------------------------------
Description:
ClusterManagementGroupManager registers a message handler for
ClusterStateMessages in its {{startAsync}} method. This approach can result in
a race, when a ClusterStateMessage is missed, because it was received before
the handler was registered.
As a solution, we can register the handler in the constructor, but this handler
should also wait for the ClusterManagementGroupManager to start before being
able to process new messages.
> Nodes can miss ClusterStateMessages on start
> --------------------------------------------
>
> Key: IGNITE-22180
> URL: https://issues.apache.org/jira/browse/IGNITE-22180
> Project: Ignite
> Issue Type: Bug
> Reporter: Aleksandr Polovtsev
> Priority: Major
> Labels: ignite-3
>
> ClusterManagementGroupManager registers a message handler for
> ClusterStateMessages in its {{startAsync}} method. This approach can result
> in a race, when a ClusterStateMessage is missed, because it was received
> before the handler was registered.
> As a solution, we can register the handler in the constructor, but this
> handler should also wait for the ClusterManagementGroupManager to start
> before being able to process new messages.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
