[ 
https://issues.apache.org/jira/browse/IGNITE-23741?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17922335#comment-17922335
 ] 

Aleksandr Polovtsev commented on IGNITE-23741:
----------------------------------------------

I'd like to clarify that the description of this ticket is a little bit 
misleading: point 1-4 are correct, however `destroyPartitionStorages()` never 
gets called, because it is called under a busy lock. On node stop, TableManager 
gets stopped earlier than PartitionLifecycleManager, so this lock is locked and 
a NodeStoppedException gets thrown and storages don't get destroyed. This still 
requires fixing, though

> Partition storages are destroyed on zone partition stop
> -------------------------------------------------------
>
>                 Key: IGNITE-23741
>                 URL: https://issues.apache.org/jira/browse/IGNITE-23741
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Roman Puchkovskiy
>            Assignee: Mikhail Efremov
>            Priority: Major
>              Labels: ignite-3
>
> When PartitionReplicaLifecycleManager is stopped (on node stop), the 
> following call chain is produced:
>  # PartitionReplicaLifecycleManager#cleanUpPartitionsResources()
>  # PartitionReplicaLifecycleManager#stopPartition()
>  # AFTER_REPLICA_STOPPED event
>  # TableManager#onZoneReplicaStopped()
>  # destroyPartitionStorages()
> So an innocuous node stop will cause data destruction, as it seems.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to