OAS Load Balancing on multiple JVMs doesn't work properly

Re: OAS Load Balancing on multiple JVMs doesn\'t work properly

Gotcha. When you set the JVM count to 4 in SM, that is equivalent to logging into the OAS EM console and setting VM count to 4 (Administration/Server Properties/Multiple VM Configuration/Number of VM Processes). My curiosity has been sated. I thought about it some more after inquiring, and realized there was probably no better way for you to be doing this in a "simple configuration".
 
Re: OAS Load Balancing on multiple JVMs doesn\'t work properly

Indeed. In fact it was so easy compared to Websphere clustering, which admittedly isn't that hard, that I figured I had got it wrong!

In many ways OAS is so much easier than WebSphere its insane. Which is the reason I went that route.

Interestingly eough I recently did a Peoplesoft course, and weblogic was similarly easy compared to OAS on that product.
 
Re: OAS Load Balancing on multiple JVMs doesn\'t work properly

Here is something interesting for you to look at. If you add the following stanza to the end of your mod_oc4j.conf file

------------------------------

<Location /oc4j-service>
SetHandler oc4j-service-handler
Order deny,allow
Deny from all
Allow from localhost
</Location>

-------------------------------

You can inspect the OC4J routing in real time from the server where OAS is installed.

It gives you a little bit of insight into what is going on behind the scenes. It can also help if you're attempting to modify the default behavior of the OC4J routing and want to confirm that it is actually working. I hope you or someone can see the value in that and/or find it potentially useful.

You can, optionally, add your hostname or ip to the allowed list and then navigate to the URL provided below the stanza.

Restart HTTP_Server after making the change to mod_oc4j.conf, then navigate to the oc4j-service status URL:

http://yourhostnamehere/oc4j-service?cmd=p

With this URL, you'll get something like this:


mod_oc4j Global Configuration

hostname : yourhostnamehere
local instance : oasinstall.yourhostnamehere
select method : Round-Robin
select affinity : None
routing mode : Static-Dynamic
routing ID : g_rt_id

OC4J Dynamic Routing

application : ascontrol
context : /em
process (Jgroup): 0

application : datatags
context : /webapp
process (Jgroup): 0, 1, 2, 3

application : WSIL-App
context : /inspection.wsil
process (Jgroup): 0

application : default
context : /jmsrouter, /j2ee, /.rmiTunnel
process (Jgroup): 0, 1, 2, 3

application : system
context : /dmsoc4j, /JMXSoapAdapter
process (Jgroup): 0, 1, 2, 3

application : JDE
context : /jde
process (Jgroup): 1, 2, 3



OC4J Static Routing

destination : JDE
context : /jde, /jde/*
process : 1, 2, 3



IASPT Process List



OC4J Process List

process,host,port,is_ssl,intance,island,ias instance name,opmnid,status,ref/max,metric,errors
0 : yourhostnamehere, 8888, 0, home, default_group, oasinstall.yourhostnamehere, 1847750767, ALIVE, 0/0, 50, 0
1 : yourhostnamehere, 12507, 0, JDE, default_group, oasinstall.yourhostnamehere, 1847750771, ALIVE, 0/0, 50, 0
2 : yourhostnamehere, 12508, 0, JDE, default_group, oasinstall.yourhostnamehere, 1847750772, ALIVE, 0/0, 50, 0
3 : yourhostnamehere, 12502, 0, JDE, default_group, oasinstall.yourhostnamehere, 1847750769, ALIVE, 0/0, 50, 0


I have sort of preached its simplicity for a couple of years. I don't necessarily favor one or the other, I just happen to support OAS so I need to know just enough about it to get by.

I think there is some obvious trade-off in simplicity versus the ability to make more granular changes, but OAS is still complex. EM and SM just add layers on top, in an attempt to ease some of the manageability headaches.

I'm looking forward to either a merged OAS/Weblogic product (Weblogic Lite?) or a clear direction from Oracle that they are going to either support both, or dump OAS in favor of Weblogic. Whatever happens, I would just like to see something happen.
 
Re: OAS Load Balancing on multiple JVMs doesn\'t work properly

Just an update.

I applied all the fixes, reset the heap size etc and it didn't make much difference.

So this morning I cracked, and just upped the JVM to the max of 6.

Since then its run 6 or 7 user per JVM on one of the machines and the other is starting to look better...
 
Re: OAS Load Balancing on multiple JVMs doesn\'t work properly

I've now fixed this by:

1) Adding 'Terry's' fixes
2) Reducing the session timeout a bit
3) Reducing the grid data timeout a bit
4) Reducing the initial JVM size from 512 to 256
5) Reducing the JVM count from 3 to 4

All Java memory errors have gone, and the three JVMs are balancing perfectly, the maximum user load difference I've seem now is 2...
 
Back
Top