Hi,
We are currently planning to go from datastage version 8.0.1(fixpack 2 )to 8.1 (on solaris).
Now I've consulted the IBM website and instead of doing a fresh install and recreate every project (and reimport the jobs), it is also possible to upgrade.
Does anyone know what the best course of action is ? Upgrade or do a fresh install ?
Something I did not find in the documentation, if you do an upgrade ,do you need to recompile all the jobs or is that step taken care of during the upgrade process ?
Has anyone done an upgrade recently and encountered serious problems ?
It's the should part that worries me :D
We've seen alot of problems in 8.0.1 moving from one fixpack to the other (and some additional loose patches) before we had a stable production environment.
For the project plan I need to asses if we need to do a clean install or upgrade. We have alot of jobs (+- 5000) so upgrade would probably be faster (due to no recompilation needed), but it also sais that the throughput time of the upgrade is dependent on the repository.
Do you happen to know if there are any benchmarks out there pointing out how long an upgrade for eg 1000 jobs would take ?
Of course, there's been plenty of times when I've ended up covered with should.
I've done a ton in the 7.x world but never one in 8.x hence my statement. Hopefully someone who has done this will wander by and help. And it should just take a matter of minutes to upgrade 1000 jobs, that's not the hard part of the process.
-craig
"You can never have too many knives" -- Logan Nine Fingers
I have no idea either. If you were upgrading from 7 to 8 the answer would be CLEAN INSTALL ONLY, but I'm not certain for an 8.0 to 8.1 upgrade.
I refused to take any of my clients to 8.0, it was just too buggy, so I've not had to deal with upgrading any of them yet (I moved them to 8.1 when it came out).
I will say this...
1) Clean install is almost always safer - no matter what product / environment
2) Clean install is almost always more work...
3) No matter what - I always recompile everything. It only takes an hour or two and brings your object code up to the current level.
The unfortunate problem is I don't know if the effort is justified because I cannot quantify the difficulties an upgrade in place would cause.
That's been my "problem" as well, the direct skipping of the 8.0 release and all of its glory. And ah... forgot about the PX compiler, especially when transformers are involved. Server stuff is automagic, PX stuff perhaps... not so much.
-craig
"You can never have too many knives" -- Logan Nine Fingers
If you are in a production environment then I would schedule for the additional work that a new install entails and do that. Particularly at V8 with the new repository structures I feel more comfortable starting with a clean slate. If you are still in development only then I'd start off with an update. Should things go wrong then you'll find out sooner rather than later and then you'll get to use one of those 4 or 5 backup copies you made of the original DataStage installation and system(s).
Ok so far nothing concrete Well we got unix engineers to agree to set up a lab environment in which we will do a proof of concept. Depending on the issues we get, it will be either clean install or upgrade
Will try to keep you up to date.
OK, I'll be concrete. I have done three 8.0.1 to 8.1, and always done a clean install. This was driven by mistrust of the internal structures of 8.0.1 some of which, at least, have to responsible for some of the bugs. If you have the choice, prefer 8.1.2 as there are more desirable features.
IBM Software Services Group
Any contribution to this forum is my own opinion and does not necessarily reflect any position that IBM may hold.
I like doing a clean install, but I don't have any experience with version 8.0. However, we have been very happy with the latest release of DataStage. I am sure we will find a few things here and there, but so far we have found the latest release to be excellent. I hope that when you have completed your upgrade that you experience similar results.
</a>