DSXchange

A question for the recent upgraders out there - is the most current version of DataStage Server 'stable'? I seem to remember seeing posts here and on ADN where people were having numerous problems after upgrading to 7.5.1. I believe patches were made available to solve some (but not all?) of the issues. Ring any bells?

We are looking at upgrading our HP/UX servers from 7.0.1 to 7.5.something. I did some testing months ago with 7.5 and only had a small problem (eventually patched) with the Notification Activity stage only working for the Admin user. I'm wondering if we should just stick with the original 7.5 release or take our chances with 7.5.1A which is listed as the latest version for HP/UX.

Thanks!

I would class it as usable with most of the problems in the new 7.5 and 7.5.1 features. Most of our problems are on the client designer side and not on the server side.

There is the 100% CPU utilisation you sometimes get in Designer when you have a hanging link open (which I think is linked to the new 7.5 visual hints feature) which only happens on some PCs for some logins.

None of us could get the new multi client manager working as it runs as a Windows service and it was very tempremental at the installation stage when demanding a user name and login, it rejected every user name on our machine. Again this is a new 7.5 "feature".

There is the extremely annoying backwards functionality of sequence jobs in that you can no longer copy a job activity stage, change the job name and retain job parameter settings. Under 7.5.1 it clears those settings out and you have to manually set parameters for each job activity stage you need.

We also found that 5 seat developer license is not enough for four developers, and sometimes it is not enough for 2 developers. Having multiple managers and designers open at the same time is taking up more then 1 license per PC. We have to have two of our developers installed under the production client license and 2 under the development license just to be able to work.

These are all design time quibbles. Can't say I've seen any problems with running jobs, certainly no spontaneous aborts, so on the server side I would judge it more stable then other 7.x versions and on the client side less stable.

There are some 7.5.1 new features that are good and some that are bad (up to you whether you can do without them):
- the SQL Builder is annoying
- the job report enhancement that links a job image to job properties is great and I didn't know it was there until I read the readme buried on the CD
- the new stored procedure stage may excite some people
- HTML job reporting is good
- globally setting annotation properties is good
- dsjob is now available on the client machine and would be great if it didn't fail on all our PCs.
- Job Sequencer looping is great but a bit buggy (often have to delete and add end loop stages to avoid compilation errors)
- thread based monitoring in parallel jobs is very good
- the new parallel lookup stage has a much better interface and the conditional lookup is okay (though it doesn't work in sparse mode where it would be even more useful then normal mode).

We went from 7.1 to 7.5.1 for the parallel Message Handler. Hands up everyone out there running Enterprise Edition with warning messages in all their jobs? We found important warnings were getting lost in all the noise. We cleaned up our jobs to reduce the number of warnings and used the message handler to remove the others, so now any new warning that appears will get noticed.

Wow, thanks Vincent. Wasn't expecting anything quite so... complete.

Some of the Client side quibbles could be rather major for us - especially the apparent change in the license tracking. Is this acknowledged as a bug?

I recently had to install the MCM on my work laptop for a PX class, needing the 7.0.1 and 7.5 Client. I had the same issue with it rejecting any username during installation. I don't recall exactly what I did to get past that part... left it blank, perhaps... but I did end up with it working for me. Would have been rather painful if it hadn't.

By 'Parallel Message Handler' is that the ability to pick specific messages from the log and 'promote' or 'demote' them, severity wise?

Yes, we use the message handler to filter out known warnings that do not lead to loss of rows or data corruption. You get a few of these in parallel jobs. In some stages you get a warning for an event that is not a problem at all. See the message handler FAQ. Warning messages were being completely ignored by production support because they were happening for every job. We are introducing a project wide standard of zero warning jobs and support have to investigate every new production warning message.