[
https://issues.apache.org/jira/browse/IGNITE-4648?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16014143#comment-16014143
]
Yakov Zhdanov commented on IGNITE-4648:
---------------------------------------
Dmitry, can you please add a test that fails before the fix and passes after?
In Ignite we try to react on a problem reported with a test that fails. Then
we fix the test.
Please look at this test -
org.apache.ignite.internal.processors.cache.AbstractCacheJtaSelfTest
You can add new test method and do some putAsync() for keys that are currently
locked by pessimistic transactions in parallel threads. Then you commit JTA
transaction and then commit cache transaction in parallel thread and then
assert the results.
Thanks!
Yakov
> IgniteInternalTx.prepare() does not wait for async operations to complete
> -------------------------------------------------------------------------
>
> Key: IGNITE-4648
> URL: https://issues.apache.org/jira/browse/IGNITE-4648
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Affects Versions: 1.7
> Reporter: Pavel Tupitsyn
> Assignee: Ryabov Dmitrii
> Priority: Minor
> Fix For: 2.1
>
>
> {{commit}} and {{rollback}} wait for async operations by calling
> {{tx.txState().awaitLastFut();}} (see {{GridCacheSharedContext}}).
> There is no such thing in {{IgniteInternalTx.prepare()}} implementations.
> Since {{prepare}} is an internal method, this is not an issue mostly, except
> for two things:
> * JTA. {{CacheJtaResource}} calls {{prepare()}} explicitly.
> * .NET {{TransactionScope}} API. Same thing as JTA, basically.
> {{PlatformTransactions}} call {{prepare()}} as well.
> As a result, if user starts an async operation within JTA transaction and
> then completes the tx, undefined behavior is possible.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
