[
https://issues.apache.org/jira/browse/IGNITE-4931?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16149227#comment-16149227
]
Alexey Kuznetsov commented on IGNITE-4931:
------------------------------------------
[~DmitriyGovorukhin] There is a problem with deactivation while lock is held
(in test _ClusterStatePartitionedSelfTest.testDeactivationWithPendingLock_).
Lock is held on grid0, when we start deactivating, grid0 hangs waiting for lock
to be released(in _waitPartitionRelease()_ ). It means, grid 0 doesn't send
GridDhtPartitionsSingleMessage so cache deactivation doesn't end properly.
I suppose, firstly we should fix waiting on local explicit lock(e.g. ignore
waiting the lock if it is a deactivation exchange future) . Do you agree ? If
so, should i create a separate ticket for this ?
> Safe way for deactivate cluster
> -------------------------------
>
> Key: IGNITE-4931
> URL: https://issues.apache.org/jira/browse/IGNITE-4931
> Project: Ignite
> Issue Type: Task
> Components: general
> Affects Versions: 2.0
> Reporter: Dmitriy Govorukhin
> Assignee: Alexey Kuznetsov
> Labels: MakeTeamcityGreenAgain, Muted_test
> Fix For: 2.3
>
>
> We must provide safe way for deactivate cluster, i mean we must wait while
> all cache operation, transaction and etc. comleted before start deactivation
> process, in current implementation we do not wait while transaction comlete,
> (forcibly stop cache during transaction).
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
