[
https://issues.apache.org/jira/browse/IGNITE-10092?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16685350#comment-16685350
]
Vladislav Pyatkov commented on IGNITE-10092:
--------------------------------------------
That does not relevant to current master.
After fixed IGNITE-9420 restorePartitionStates performed fully into cplock.
> Race in partition state when checkpoint started at the middle of starts caches
> ------------------------------------------------------------------------------
>
> Key: IGNITE-10092
> URL: https://issues.apache.org/jira/browse/IGNITE-10092
> Project: Ignite
> Issue Type: Bug
> Reporter: Vladislav Pyatkov
> Assignee: Vladislav Pyatkov
> Priority: Major
>
> In the method {{GridCacheDatabaseSharedManager#restorePartitionStates}}
> called method of creation partition at start
> {{GridDhtPartitionTopology#forceCreatePartition.}}
> Partitions create in MOVING state, and can be accessible to on checkpoint
> thread.
> By the reason checkpoint thread
> ({{GridCacheOffheapManager#saveStoreMetadata}}) can to read MOVING partition
> state and store it after
> {{GridCacheDatabaseSharedManager#restorePartitionStates}} completion in
> parallel.
> In result partition stet stored incorrect, should leads to rebalance.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
