[
https://issues.apache.org/jira/browse/IGNITE-1349?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14727306#comment-14727306
]
Semen Boikov commented on IGNITE-1349:
--------------------------------------
Found reason for the hang: it is possible that when method
'GridCacheMvccManager.addNext(GridCacheContext cacheCtx, GridCacheMvccCandidate
cand)' is called from system pool thread then thread-local collection 'pending'
has not cleaned entries related to another transaction.
In GridNearOptimisticTxPrepareFuture.prepareOnTopology: after waiting on
topology future it can execute 'prepareOnTopology' from system pool, but does
not clean thread context after this.
Need to check all code related to transactions for similar issues.
> CachePutAllFailoverTxTest hangs
> -------------------------------
>
> Key: IGNITE-1349
> URL: https://issues.apache.org/jira/browse/IGNITE-1349
> Project: Ignite
> Issue Type: Sub-task
> Components: cache
> Affects Versions: ignite-1.4
> Reporter: Semen Boikov
> Assignee: Semen Boikov
> Priority: Blocker
> Fix For: ignite-1.4
>
>
> Test CachePutAllFailoverTxTest executes cache.putAll concurrently with node
> start/stop. After some time it fails because of there are some transactions
> which can not finish prepare step. Hang reproduces if size of map for putAll
> is large ehough (~100).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
