Using Subversion as source code mgr with Information Server

deboots · Post by **deboots** » Tue Aug 28, 2012 11:29 am

Hi All - We need to implement version control for our DataStage jobs and we are looking at Subversion. There are several older posts on DSXchange indicating success with Subversion. We have it installed and working - but wondering about best options for the actual process on top of the technology. If you have implemented Subversion (SVN) with InfoSphere 8.5 or 8.7 (ideal) would you mind sending info on your process for managing code? For example, check out doesn't seem to lock the code - meaning multiple developers could checkout the same job and then check it back in not realizing they were over-writing each other. Allowing Developers to 'lock' the job in Production as they check it out requires write access which opens us up to accidental deletes - not good. It makes sense to centralize/manage the move to prod step but we would like not to centralize the check out (so much overhead). How have you set this up? Dev and Prod environments with Dev wide open and Prod locked down to SysAdmin / Version Control manager? Any tips? Can I buy you a cup of coffee (or bring you a Pacific NW beer) at IOD for helping me out?? Happy to share any of our process documentation in return!

vmcburney · Post by **vmcburney** » Tue Aug 28, 2012 4:59 pm

Check out and check in is not crucial in DataStage because there is only one live copy of the code. It's not like a scripting language where two people can check out code and be working on two different copies. Generally version control is used via the Information Server Manager at the time that code is migrated between environments - it is part of the release management and not part of developer management. It is used to baseline versions of DataStage jobs as they move from development to test to production.

Most DataStage developers are using DataStage Designer for their work and hardly ever open up IS Manager - so they tend to ignore check in check out altogether. I think you can track developer changes to jobs via Metadata Workbench queries and save the check in check out for the release manager. Eventually IBM need to integrate source control into the DataStage Designer but until they do that it's hard to set up water tight processes.

ray.wurlod · Post by **ray.wurlod** » Tue Aug 28, 2012 5:12 pm

That said, using Information Server Manager and the Eclipse plug-in for SubVersion you can quite easily set up "proper" source code control for DataStage. But there is, as Vincent suggested, an education/management component to it as well - experienced DataStage users will need to be convinced (!) to go the check out/check in route.

PaulVL · Post by **PaulVL** » Wed Aug 29, 2012 8:42 am

Our shop does not compile in PROD, so we use an external source control system.

We tell (force) the user to extract their DSX or ISX files and have the source and executable in that file. Edits are done in DEV, extracted and checked into the repository, then migrated to QA. After testing it goes from repository to PROD.

NO COMPILING IN QA, NO COMPILING IN PROD.

FranklinE · Post by **FranklinE** » Mon Sep 10, 2012 10:53 am

PaulVL wrote:NO COMPILING IN QA, NO COMPILING IN PROD.

I believe this is the only way to go with version control. We import executables only in QA and Prod.

I'm working with our Subversion team to create user-requested automated functions for importing executables (and design code, but then we go a bit further than most shops and use version control from DEV on up), from the SV interface. This will eventually replace BladeLogic packages, which currently are a very large thorn in my side.

I am not able to share details around all that, regretfully, so please don't ask. I will share as much as I can as we work out the procedures.