[jboss-jira] [JBoss JIRA] (WFWIP-222) when client is scaled down with in-doubt transactions, tx participants on server are resolved with delay

[Ondrej Chaloupka (Jira)]

     [ https://issues.jboss.org/browse/WFWIP-222?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Ondrej Chaloupka closed WFWIP-222.
----------------------------------
    Resolution: Rejected


The behaviour is expected and the transaction resolution is in hands of different component than the recovery. The transactions on the server were started but was not moved to the {{PREPARED}} state. They are in-flight stored only in memory. It's responsibility of the transaction reaper to finish them. It really does so but the reaper is not dependent on the recovery processing. The reaper roll-backs the transactions after transaction timeout is over. If the reaper is not capable to do so for the remote resources (e.g. WildFly goes down meanwhile) then the database/jms/... will roll-back the transaction after the timeout on its own (aka. prepare was not called -> no promise from the remote resource on the processing was done -> transaction timeouts after the timeout is over)

> when client is scaled down with in-doubt transactions, tx participants on server are resolved with delay
> --------------------------------------------------------------------------------------------------------
>
>                 Key: WFWIP-222
>                 URL: https://issues.jboss.org/browse/WFWIP-222
>             Project: WildFly WIP
>          Issue Type: Bug
>            Reporter: Martin Simka
>            Assignee: Ondrej Chaloupka
>            Priority: Blocker
>         Attachments: tx-server.log
>
>
> this follows up on WFWIP-206
> While testing tx recovery in OpenShift I see that when client is scaled down with in-doubt transactions, tx participants on server are resolved with delay.
> Scenario:
> *ejb client* (app tx-client, pod tx-client-0):
> * EJB business method
>   ** lookup remote EJB 
>   ** enlist XA resource 1 to transaction
>   ** enlist XA resource 2 to transaction
>   ** call remote EJB
> *ejb server* (app tx-server, pod tx-server-0):
> * EJB business method
>   **  enlist XA resource 1 to transaction
>   ** enlist XA resource 2 to transaction
> *testTxStatelessClientSecondPrepareJvmHalt*
> JVM on client crashes in PREPARE phase of second XA resource on client. Then openshift restarts pod, but pod is immediately scaled down. Transactions on client pod are rollbacked during scale down, but on server they are rollbacked some time later. I'm not sure if it is periodic recovery or tx timeout.
> server log with {{com.arjuna}} trace attached. 
> Feel free to reject if it is expected behavior. 



--
This message was sent by Atlassian Jira
(v7.13.8#713008)