[
https://jira.nuxeo.com/browse/NXP-7943?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Florent Guillaume updated NXP-7943:
-----------------------------------
Affects Version/s: (was: 5.5-SNAPSHOT)
5.3 GA
> Fix Oracle tightly-coupled transaction optimization XA bug
> ----------------------------------------------------------
>
> Key: NXP-7943
> URL: https://jira.nuxeo.com/browse/NXP-7943
> Project: Nuxeo Enterprise Platform
> Issue Type: Bug
> Components: Core SQL Storage
> Affects Versions: 5.3 GA
> Reporter: Florent Guillaume
> Assignee: Florent Guillaume
> Priority: Major
> Fix For: 5.5
>
>
> When using XA with more than one resource, Oracle by default uses a
> "tightly-coupled transaction" mode for the various transaction branches of
> the resources. In this mode, all the resources except the last one return
> {{XA_RDONLY}} when {{prepare()}} is called, and only for the last resource
> does {{prepare()}} return {{XA_OK}} thus triggering the normal {{commit()}}.
> This is likely done to avoid doing one more round-trip to the database per
> resource.
> However it means that receiving {{XA_RDONLY}} from {{prepare()}} does not
> necessarily mean that no work was done by the resource, so there may still be
> invalidations that have to be sent.
> See
> http://docs.oracle.com/cd/E14072_01/appdev.112/e10471/adfns_xa.htm#ADFNS810
> for reference.
> A consequence of not sending invalidations is that data may appear
> inconsistent between two session.
> Note that this bug only happens with two or more XA resources, which can only
> be the case with two or more VCS repositories, or a VCS repository and a XA
> JDBC datasource.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
_______________________________________________
ECM-tickets mailing list
ECM-tickets@lists.nuxeo.com
http://lists.nuxeo.com/mailman/listinfo/ecm-tickets
