[infinispan-issues] [JBoss JIRA] Commented: (ISPN-994) Crete tooling (JMX) to handle heuristic transactions

Mircea Markus (JIRA) jira-events at lists.jboss.org
Mon Mar 28 06:50:37 EDT 2011


    [ https://issues.jboss.org/browse/ISPN-994?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12592101#comment-12592101 ] 

Mircea Markus commented on ISPN-994:
------------------------------------

It is rather difficult to pass the XID (as received from TM's recover process) to ISPN's JMX operations: the XID is primarily an byte[]; converting a String to a byte[] is difficult due to encodings. In order to avoid this:
- have an operation shows the mapping XID->internal value (e.g. internal value might be a long)
- the sys admin visually maps XID to the internal value
- it calls forceCommit(long)/forceRollback(long)
This is not very user friendly but still easier than coming up with a XID based version.
On top of that, the XID based operations are still exposed (i.e. forceCommit(XID)/forceRollback(XID)), as applications might want to access them programatically).

This solution requires generating a cluster wide unique long for each transaction/XID. AFAIK we have something similar for HR unique ids that can be reused.
On the top of my head a way to generate such a unique long is:
- on each node have a shared static counter that increments for each new tx (long)
- replace the most relevant decimal digit from the long with the viewId
- replace the 2nd most relevant digit from the long with the node's index


 
 



> Crete tooling (JMX) to handle heuristic transactions
> ----------------------------------------------------
>
>                 Key: ISPN-994
>                 URL: https://issues.jboss.org/browse/ISPN-994
>             Project: Infinispan
>          Issue Type: Feature Request
>          Components: Transactions
>            Reporter: Mircea Markus
>            Assignee: Mircea Markus
>             Fix For: 5.0.0.CR1, 5.0.0.FINAL
>
>
> now that ISPN-272 is implemented the TM is aware about the heuristic transactions. 
> After TM determines such a transaction it informs the System Administrator. It's SA's job to force the rollback/commit of those pending transactions.
> In order to do that he needs tooling in order to access the cluster. A solution would be JMX as described here: http://lists.jboss.org/pipermail/infinispan-dev/2011-March/007651.html
> This is complementary to ISPN-272 and very important as without it transaction failure cannot be fixed.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira


More information about the infinispan-issues mailing list