Version 8- Does it need more resources for the dev server?

dnsjain · Post by **dnsjain** » Mon Nov 03, 2008 9:43 am

We had IBM consultant at our site and he suggested that we should have eight processor server for DataStage development server and four process server for DataStage production server. His argument is because of new architecture, development server needs more resources.

I am not very convinced with his argument. I thought I should ask experts here at the forum, who has version 8 installed at their site and are using it.

Thanks,
Dinesh

ray.wurlod · Post by **ray.wurlod** » Mon Nov 03, 2008 10:37 am

Without knowing why that recommendation was made, it's impossible to comment. Can you please ask the IBM consultant why that recommendation was made? For example do you have many developers, do you use the same machine for development and testing?

dnsjain · Post by **dnsjain** » Mon Nov 03, 2008 12:36 pm

Ray,

We are going to have 10 to 15 developers, some will be onsite and some will be offshore for DataStage job development. IBM consultants argument is, each developer will be taking resources on the server and in order to provide better development environment, eight processor server is needed.

We will have a separate server for testing and production.

Thanks,
Dinesh

jdmiceli · Post by **jdmiceli** » Mon Nov 03, 2008 1:10 pm

Odd as this may seem, I am almost in agreement with Dinesh and the consultant. IBM has told us that the Development Client tools should not be run with less than 3 GB of RAM available. It is also quite apparent from our experience thus far with V8, that the Dev box needs to be very beefy due to the total number of connections and test runs that developers will be making.

Production should not be have client GUI client connections occuring in general other than Admin, so the resources are going to be primarily devoted to doing the work of production. I don't know that I would flip the configuration as was suggested, but I can certainly see making the Dev box equivalent to the Prod box if possible to handle the extra load.

So far, we are finding V8 to be a dog on performance (V7 currently runs 4 times faster for my project) and the multi-instance jobs with auto-purge of logs bug is a literal show stopper (there is a patch).

Just my 2 cents worth.

John Smith · Post by **John Smith** » Tue Nov 04, 2008 11:17 pm

jdmiceli wrote:Odd as this may seem, I am almost in agreement with Dinesh and the consultant.
.

but Dinesh is not in agreement with the consultant

mddw08 · Post by **mddw08** » Tue Dec 09, 2008 5:52 pm

Do you know the patch name for "the multi-instance jobs with auto-purge of logs bug"? We installed JR30015 patch today that according to IBM should fix the problem, but the issue still exists...

jdmiceli · Post by **jdmiceli** » Tue Dec 09, 2008 6:20 pm

There is evidently another step that has to be done as well. I don't know if you did it, but here are the instructions as plucked from an email chain between our DataStage admins:

Go to project properties, general tab, click on environment, click on user defined, right click uner details, add new cell, enter the info below.

name=DS_NO_INSTANCE_PURGING type=string prompt=no instance purge value=1

I don't know if that will help at all, but I hope it does.

mddw08 · Post by **mddw08** » Wed Dec 10, 2008 4:46 pm

Thanks for the hint. It looks like it fixed the issue - I'm getting no more errors for now. My only fear is it may accumulate something that will blow up later.

For the past 6 month, we applied a number of patches that while fixing the original issue would introduce another one or two.

jdmiceli · Post by **jdmiceli** » Wed Dec 10, 2008 7:15 pm

I can relate. That is known as the 'addition of undocumented features.'

Don't you know that is a good thing!

eostic · Post by **eostic** » Thu Dec 11, 2008 7:19 am

...to the original question, regarding size of the (engine) machine, it's often the case that the development machine may need to be bigger than the production machine. This is especially true when simultaneous developers are building, compiling and testing concurrent parallel jobs ---- Jobs that at production run-time may run serially in a Sequence. Even a single developer will often run multiple Jobs during testing that otherwise might not run concurrently. Server Jobs are no exception from this, but of course parallel Jobs will potentially spin up a lot more processes depending on the complexity of the Job and the degree of parallelism.

Ernie

asaf_arbely · Post by **asaf_arbely** » Thu Jul 09, 2009 5:37 am

jdmiceli wrote:There is evidently another step that has to be done as well. I don't know if you did it, but here are the instructions as plucked from an email chain between our DataStage admins:

Go to project properties, general tab, click on environment, click on user defined, right click uner details, add new cell, enter the info below.

name=DS_NO_INSTANCE_PURGING type=string prompt=no instance purge value=1
I don't know if that will help at all, but I hope it does.

Hi all
We also install the patch 30015v3 in order to resolve the logs disappearance
After installing we encountered several bugs -
Synchronization error
Parameter transfer between sequence and job fails
Does anyone know of a solution ?

martind · Post by **martind** » Thu Jul 09, 2009 11:01 pm

Do you intend to use other tools in the suite, such as Information Analyzer, on the dev server? This tool would typically only be used in a dev environment.