[jboss-jira] [JBoss JIRA] Resolved: (JBAS-5647) HAPartition to support wait for first positive response

[Galder Zamarreno (JIRA)]

     [ http://jira.jboss.com/jira/browse/JBAS-5647?page=all ]

Galder Zamarreno resolved JBAS-5647.
------------------------------------

    Resolution: Done

New HAPartition API implemented and upgraded trunk to
ha-server-api 1.1.0.CR2.

> HAPartition to support wait for first positive response
> -------------------------------------------------------
>
>                 Key: JBAS-5647
>                 URL: http://jira.jboss.com/jira/browse/JBAS-5647
>             Project: JBoss Application Server
>          Issue Type: Feature Request
>      Security Level: Public(Everyone can see) 
>          Components: Clustering, Naming
>            Reporter: Galder Zamarreno
>         Assigned To: Galder Zamarreno
>             Fix For: JBossAS-5.0.0.CR2
>
>         Attachments: 2008-07-12.15-44.tgz
>
>
> Create a new method in HAPartition or modify an existing one so that we can provide 
> the capability of calling a method cluster wide but waiting only until the first positive 
> response has been received.
> The definition of positive response would be defined by a RspFilter impl that tells 
> JGroups it's OK to return as soon as a valid response is received. The call would 
> still be GET_ALL.
> An example of this can be found in JBC's DataGravitationInterceptor.gravitateData() 
> where it filters NNFEs.
> RspFilter concept doesn't exist in JGroups 2.4, so this won't be implemented for 
> AS 4.x.
> The use case for this would be the following:
> Let's say you have a cluster of N nodes and you have a jar deployment in M nodes 
> where M is less than N. A request comes to a node that does not have this jar 
> deployment, so it tries to look it up cluster wide. At this point all M nodes reply with 
> the binding but one of the nodes that didn't have the binding is having GC issues so it
> does not respond in time. With the current set up, AS would have to wait for all nodes
> to reply back before it can return any of the positive responses received. The idea 
> here is for positive results to make JGroups return straight back.
> If M was 1 and this was the node with GC problems, not much we could do.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira