Hi there,

It is taking half an hour to get 1.5 MB .ear file to a Cell which has about 20 Clusters spanned across two Nodes.

I have Deployment manager with only 256MB as the JVM Heapsize, is this the constraint ?

Could any one put some light in reducing the amount of time in deploying these apps of bigger size.


Satish

I've see some slow behavior in deployment managers, but they were typically running on VMs with unreserved resources. Other times, I've found that the NIC, network activity, or the target server was experiencing issues.

Currently, my DMGRs are not configured with a Min/ Max heap size, but they average about 600 MB.

Can you elaborate more on the topology?  You mention 20 clusters, but how many nodes, are they hosted on physical or virtual hardware; do you leverage node groups; which version of ND are you running?  Also, is this a new issue or something that has been ongoing since the cell was created?

HI there,

Thanks for the response.

We are running

Websphere 6.1.0.35 on RHEL 5.4 on z/VM on z196 (Mainframe)

We have One DM with two Nodes. We have 20 clusters spanned horizontally across those two Nodes. This is a UAT environment which is  similar environment to PRODUCTION environment in terms of number of components, memory but not it terms of CPU preference.

And we deploy only .ear file at a time.

 

 

 

 

Just to add couple of bits to the previous response.

The delay in deployments were always the case.

While I'm not familiar with the mainframes, I'd start out by looking at the jvm pmi information. It might give you an idea how much CPU the process is consuming or the heap split. You can also look at the garbage collection to see if it's kicking off frequently when youre trying to deploy. Also the deployment manager should have tuning recommendations which might shed some light on the situation.

There's always the system out.log for the deployment manager which might you a starting point.

Is there a reason why you have the dmgr capped at 256 mb? If there's available ram you may want to uncap the dmgr max heap size and see if the performance increases.

If all of the above fail, you can call in pmr with IBM. I've always found their response time and resolutions to be very fast and helpful.

Good luck!

Thanks Erik

Any luck with improving the performance?

do you deploy via script or Gui ?

Thanks Erik for following up.I was engaged in some other priority stuff. Will work that out soon. Thanks.

 

 

We use scripts to deploy.

 

There are several options on AdminApp.install() that might help.
I would explore which one fits your app and use these options;
here is a sample of these options -

-usedefaultbindings
-nopreCompileJSPs
-distributeApp
-nouseMetaDataFromBinary
-nodeployejb 
-createMBeansForResources
-noreloadEnabled
-nodeployws
-validateinstall warn
-noprocessEmbeddedConfig
-noallowDispatchRemoteInclude
-noallowServiceRemoteInclude
-asyncRequestDispatchType DISABLED
-nouseAutoLink

 
also, if you have a web app in the ear you may need: -MapWebModToVH

Check the Info Center on these options as they can speed up the install.