If you are unable to create a new account, please email support@bspsoftware.com

 

TR3110 Error when running cube build twice in a row via Transformer or cmd line

Started by brian_brooks, 17 Nov 2016 02:52:43 PM

Previous topic - Next topic

brian_brooks

Hi all,
We have a somewhat strange issue with a 10.1.1 cube that has been running fine for the last few years via command line script suddenly causing problems. The only thing that has changed is the service account for the signon used to run the cube. This same service account is being used for other cubes and works perfectly fine.

The strange issue is this: we can open the cube model in Transformer, run the cube and it will build the cube successfully. If we close out of the model and reopen within Transformer, we can then run the cube to build it successfully once more so long as the model has been closed after it is built.

If we run the cube so that it builds successfully and then try to immediately run the cube to build again, it will fail with a TR3110 error message. This then places the model in a suspended state and disallows the ability to run the model any further until we replace the model with a backup.

When trying to run this via command line, it works pretty much the same way. I can run the script below once and it will build the cube successfully. If I try and run the script again, it will look like the script ran fine, but looking at the background processes will show that a transda.exe process is still lingering taking up 25% CPU.

Has anyone out there experienced this issue? We're looking to get a backup copy of the model from a couple of weeks ago to see if that might resolve the issue, but I don't have much faith in that. Here is the script used to process the file:

cogtr -n2 -r4 -dLogFileName=Regen.log -dLogFileDirectory=E:\CognosData\Production\Logs -k"NIBBLE Pork=pdxx/xxxx" -s -g -p"E:\Cognosdata\production\hps\Cube Models\C8_BGF Weekly Production Data.pyj"

bdbits

Not really sure, just throwing some ideas out there.

* Does the new account have the very same permissions as th old (I would think the first build would fail if not, but since it is not working...)? Filesystem permissions also matter.
* Does the new service account have a valid signon in the model?
* If the old account still exists, can you try it using that?
* Is there any further information in the log file the build process produces? From your command line, these would be found in E:\CognosData\Production\Logs.
* If your model is in pyj format, you might try saving as mdl and running with that file.
* What if you disable multiprocessing? (Since you indicate transda is running and I believe that is only invoked with multiprocessing.)

brian_brooks

Thanks for the reply, bdbits. We had our network admin restore a backup copy of the cube model from a couple of weeks ago, changed the signon account to the new account and everything works. Due to the shaky nature of Transformer, I'm just not going to question it further and thank the lucky stars that it worked in my favor this time!

bdbits

I know what you mean. Unfortunately, I think IBM has all but mothballed any work on Transformer aside from mostly empty version number increments and logo updates.

Glad you got it going.