Hello,
I have a question about quiescing in GlassFish 2.1, I'm following the
guide located at
http://docs.sun.com/app/docs/doc/820-4341/abdik?a=view
and am in the "Upgrading In a Single Cluster" scenario.
My GlassFish setup is pretty straightforward:
- One cluster named "testcluster"
- 2 node agents (node-app1 and node-app2)
- 2 instances (one on each node agent called instance1 and instance2)
- I deployed the sample "clusterjsp" application provided with GlassFish
- One HTTP load balancer configured in GlassFish with working auto-apply
The quiescing process works fine in itself, to deploy a new application
version without interruption I'm doing this:
1) Turn off dynamic reconfiguration:
asadmin set testcluster-config.dynamic-reconfiguration-enabled=false
2) Redeploy the new application to the server via the Web UI
3) Enable the new application for all the instances:
asadmin enable-http-lb-application --name clusterjsp testcluster
4) Quiesce one instance:
asadmin disable-http-lb-server --timeout=1 instance1
5) Stop/start the instance to deploy the new application
asadmin stop-instance instance1
asadmin start-instance instance1
6) Re-enable the instance on the load balancer
asadmin enable-http-lb-server instance1
7) Repeat steps 4, 5 and 6 for instance 2
Now my problem is that as soon as I run any of the
"disable-http-lb-server" or "enable-http-lb-server" commands on ANY of
my instances, all other instances get a "Restart Required" status in the
GlassFiwh Web UI. This defeats the point of a rolling upgrade quite a bit :)
Do I really need to restart the instances (which would render the whole
procedure useless) or is it only a glitch in GlassFish 2.1? It seems
related to "dynamic reconfiguration" being disabled but i can't see at
which level...any clue on what's happening?
PS: a picture of the end result after the 2 instances have been trough
the process described before:
http://www.alphatek.info/divers/restartrequired.png
Thanks!
Steven
