[
https://issues.apache.org/jira/browse/IGNITE-29233?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18143859#comment-18143859
]
Ignite TC Bot commented on IGNITE-29233:
----------------------------------------
Fixed explicit locks that could remain acquired after a tryLock timeout during
mapping and block partition map exchange.
The Release Note field has been filled with the wording above. Please review it
and edit it if needed.
> Fix explicit lock left acquired when its timeout expires during mapping
> -----------------------------------------------------------------------
>
> Key: IGNITE-29233
> URL: https://issues.apache.org/jira/browse/IGNITE-29233
> Project: Ignite
> Issue Type: Bug
> Reporter: Nikita Amelchev
> Assignee: Nikita Amelchev
> Priority: Critical
> Fix For: 2.20
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Fix explicit lock left acquired when its timeout expires during mapping
> If the timeout of {{IgniteCache.lock(key).tryLock(timeout)}} expires while
> the lock request is still being mapped (a GC pause longer than the timeout,
> or the mapping waits for an exchange), the lock future is completed and
> undone *before* its candidates are registered. {{tryLock}} returns {{false}},
> but:
> * the key stays locked on the primary node with no thread able to release it;
> * the next partition map exchange hangs waiting for the leaked candidate.
> A reentry of a key the thread already holds is not reverted either when a
> multi-key lock times out, so the key stays locked after the holder releases
> it.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
