Hey,
a) don't share things between cluster nodes!
b) before you change your lib at one node please shutdown the context
with manager app or ant tasks.
c) Release change at runtime with cluster is heavy and can not made
with the same nodes.
OK, when no class changed that used inside session replication, it
can be work!
d) Don't change all nodes at the same time.
Change one and
after restart context - see that all session comes in.
Then made change the others.
e) Build cluster domains and not running all nodes inside one cluster.
Tomcat cluster replicated all sessions between every node and
this is very memory intensive.
Peter
Davide Romanini schrieb:
Hi,
I'm developing some web services using Axis. The whole axis context
(containing all my services) is embedded as a servlet into Tomcat 5.5.9.
Any time I create a new version of one service, I need to replace the
previous jar file into WEB-INF/lib, so to make changes effective I also
must reload the axis context (using manager facility).
My problem is that in production environment we have 6 machines with a
shared filesystem (gpfs) in load-balancing. So I think that if I do a
reload of the context using manager, I actually reload just one of the 6
istances of Tomcat (the one answered me in that moment), but the other 5
don't see the changes made to WEB-INF/lib and continue to run and old
version of the service.
Take into account that I don't have access to each single machine of the
cluster, so I can't do a reload of all the 6 contexts from my machine.
There's a way to propagate a context reload between the clusters?? I
really need to access each machine and reload for each single Tomcat
instance?
Thanks in advance.
Bye,
Davide Romanini
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
