[ 
https://issues.apache.org/jira/browse/OOZIE-3134?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16354017#comment-16354017
 ] 

Rohini Palaniswamy commented on OOZIE-3134:
-------------------------------------------

There is more to this issue. Please refer OOZIE-1980. 

There are two issues to be addressed.

1) Include the SLA inserts as part of the coord action table inserts and make 
them one transaction which JPA can rollback on database failure. Currently they 
are two separate JPA transactions.
2) Remove the SLA entry from the in memory map on JPA rollback.

> Potential inconsistency between the in-memory SLA map and the Oozie database
> ----------------------------------------------------------------------------
>
>                 Key: OOZIE-3134
>                 URL: https://issues.apache.org/jira/browse/OOZIE-3134
>             Project: Oozie
>          Issue Type: Bug
>            Reporter: Attila Sasvari
>            Assignee: Julia Kinga Marton
>            Priority: Major
>             Fix For: 5.0.0
>
>         Attachments: OOZIE-3134-001.patch, OOZIE-3134-002.patch
>
>
> Upon {{SLACalculatorMemory.addRegistration}}, Oozie puts an entry into an 
> in-memory concurrent hashmap ("slaMap"), and later [persists 
> it|https://github.com/apache/oozie/blob/50f4b5984832941f1341586e43fd832c293b3275/core/src/main/java/org/apache/oozie/sla/SLACalculatorMemory.java#L380]
>  in the Oozie database. 
> However, if there is a failure during the database operation, a 
> {{JPAExecutorException}} is thrown, and the entry [is not removed from the 
> SLA map| 
> https://github.com/apache/oozie/blob/50f4b5984832941f1341586e43fd832c293b3275/core/src/main/java/org/apache/oozie/sla/SLACalculatorMemory.java#L393].
> It may introduce inconsistency between the Oozie database and the SLA map.
> To prevent this, a rollback mechanism (with proper logging) should be 
> implemented.  It would also make sense to do more sanity/consistency check in 
> the Oozie server.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to