[infinispan-issues] [JBoss JIRA] (ISPN-1556) An operation that fails but it's silent shouldn't abort the transaction internally

[Galder Zamarreño (Updated) (JIRA)]

     [ https://issues.jboss.org/browse/ISPN-1556?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Galder Zamarreño updated ISPN-1556:
-----------------------------------

          Summary: An operation that fails but it's silent shouldn't abort the transaction internally  (was: Ability to test and acquire lock if available, without aborting the txn if not available)
       Issue Type: Bug  (was: Feature Request)
    Fix Version/s: 5.1.0.CR2
         Priority: Critical  (was: Major)


Imagine 2nd level cache use case. A put for external read that fails due to a concurrent load of data in the database shouldn't force the transaction to rollback. The point of put for external read is that with fail silently, it's assuming it can fail but it's fine and should not have an impact in the transaction outcome.
                
> An operation that fails but it's silent shouldn't abort the transaction internally
> ----------------------------------------------------------------------------------
>
>                 Key: ISPN-1556
>                 URL: https://issues.jboss.org/browse/ISPN-1556
>             Project: Infinispan
>          Issue Type: Bug
>    Affects Versions: 5.1.0.BETA2
>         Environment: AS7.1.0.alpha2
>            Reporter: Gary Brown
>            Assignee: Galder Zamarreño
>            Priority: Critical
>             Fix For: 5.1.0.CR2
>
>         Attachments: CacheTxStackTrace.log, Sample2Test.java, SampleTest.java, serverISPNTrace.log.zip
>
>
> I have a system that performs a large number of tasks in a single transaction for efficiency. Some of those tasks access infinispan caches.
> I found that occasionally I have been getting lock timeouts for the default 15 second period.
> Lock contention is not a problem - but the impact of failing to obtain the lock results in the whole transaction being aborted, which aborts the work also carried out for potentially a large number of other tasks, resulting in all of the work being retried.
> I was wondering if it would be possible to provide an alternative lock implementation the AdvancedCache that allowed a client app to test whether the lock was available and acquire it - returning 'true', but if the lock was not available, simply returning false, allowing the client code to make a decision about how to proceed.
> In my case, I could then add the specific task to a retry queue, and move onto the next task, committing all of the work that had been successfully completed for the other tasks.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira