Hi Daniel,
Please see inlines
glassfish_at_javadesktop.org wrote:
> Hi
>
> I have setup a cluster of 2 nodes with GF build 46.
> The cluster is configured to use in mem replication and has apache+mod_jk in front (glassfish ajp listener enabled). Doing some testing on the availability aspect I did the following with a test application (web app + Statefull SB):
>
> - Start Cluster with 2 nodes
> - access app to create http session and sfsb instances, populate sessions with some data
> - kill one cluster node -> failover to other node is working great, session data is there
> - restart the killed node, wait some seconds and access it (through lb)-> ups, session is not there!
>
> So the question is: if a failed cluster node gets restarted does it get the session data replicated back from the other instance(s) using in memory replication? Why doesn't this happen in my scenario?
>
>
The scenario is supposed to work (It does work with our tests, though
there is no mod_jk in our environment)
Just to clarify steps:
1. 2 instance cluster, access through a loadbalancer
2. Request goes to instance1
3. Stop instance 1, the next request of same session goes to instance 2
4. Start instance 1, the next request will still go to instance 2
(sticky ness)
5. Stop instance 2, next request is supposed to go to instance 1
Please let me know if you are trying above scenario
Thanks,
Gopal
> Thanks for hints
> Daniel
> [Message sent by forum member 'barz26' (barz26)]
>
> http://forums.java.net/jive/thread.jspa?messageID=216510
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: users-help_at_glassfish.dev.java.net
>
>