[jboss-jira] [JBoss JIRA] Commented: (JBAS-5197) Add "RetryInterceptor" functionality to all clustered proxies

[Brian Stansberry (JIRA)]

    [ https://jira.jboss.org/jira/browse/JBAS-5197?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12521869#action_12521869 ] 

Brian Stansberry commented on JBAS-5197:
----------------------------------------

<bstansberry> dmlloyd: i don't recall if we talked about this the other day: http://community.jboss.org/wiki/RetryInterceptor
<bstansberry> if yes i won't natter on
<dmlloyd> no we didn't cover it
<dmlloyd> well, we brushed on the existence of the problem
<dmlloyd> that's about it
<bstansberry> it was quite important w/ RMI-based invocation, since a cluster restart would mean any RMI stubs on the client were worthless
<bstansberry> less so w/ other invokers
<bstansberry> basically if the proxy couldn't invoke any servers it would try a JNDI lookup of a special location to see if it could get a new set of targets to invoke on
<dmlloyd> we wouldn't need to do that for other transport types I don't think - however, the idea of retrying (or maybe blocking) until the target becomes available again might be appealing if it can be made to work
<bstansberry> yeah.
<bstansberry> the retrying can be done in the clustering logic itself. right now it's dumb -- if contacting a target fails it never tries again
<bstansberry> it should be smarter
<bstansberry> maybe mark the target suspect, so it's not chosen in load balancing decisions (until server indicates that node s/b ok) but can still be used as an option if all others are dead
<dmlloyd> yeah, or some way of ranking targets or something
<bstansberry> yep. so there's probably no point to a JNDI lookup; the client already has all the useful info; remoting URLs don't "break" like RMI stubs

> Add "RetryInterceptor" functionality to all clustered proxies
> -------------------------------------------------------------
>
>                 Key: JBAS-5197
>                 URL: https://jira.jboss.org/jira/browse/JBAS-5197
>             Project: JBoss Application Server
>          Issue Type: Feature Request
>      Security Level: Public(Everyone can see) 
>          Components: Clustering
>            Reporter: Brian Stansberry
>            Assignee: Brian Stansberry
>             Fix For: JBossAS-6.0.0.CR1
>
>
> Container issue for ensuring "RetryInterceptor" functionality is added to all clustered proxies.
> "RetryInterceptor" functionality consists of adding an interceptor that is able to access the cluster (currently via JNDI) and get an updated set of cluster targets when all current targets have failed.  Needed when a complete cluster restart has occurred since the current list of targets was deserialized on the client.
> We currently have this for EJB2 proxies; need it for EJB3, HA-JNDI and generic AOP-based and ProxyFactoryHA-based proxies.

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