TM1 crash: deleting the subset that is the data source of a TI process

[Wim Gielis]

Hello all,

Happy New year !

But 2017 starts with a crash in TI.

For years and years, I use a TI process that:
1. Create a temporary dimension in the Prolog that and populate it (DimensionElementInsertDirect and the Update function)
2. Assign the ALL subset on that temporary dimension as the datasource of a TI process
3. Have some code in the Metadata tab, unrelated to the temporary dimension
4. Delete the temporary dimension again as the last line in the Epilog

This used to work fine. In the latest version of TM1 (Planning Analytics Local) the TI process crashes.
Without deleting the dimension in the Epilog, the process runs successfully.

Of course as a workaround I can dump the contents of the temporary dimension in a text file and use the text file as the data source, but that should not be needed.

Any ideas ?

Thanks !

Wim

[Steve Rowe]

Have you recently changed versions?

I'm wondering if there is a related in some way to this (http://www.tm1forum.com/viewtopic.php?f ... 552#p63552), where it seems like changes to a dimension are not fully committed at the end of the prolog (I am pretty sure it used to work like this).

I got around my problem by doing my build work in a distinct TI that I called from the prolog of the main TI. Since the sub-TI was fully complete within the prolog of the main TI the main TI was quite happy to talk to those objects.

HNY!

[Wim Gielis]

Thank you Steve and happy new year too !

This same process works fine in TM1 10.2.2 FP6 and not anymore in the new version.
I did not use a separate TI process for cleaning up (I could do that, or use a text file as the data source),
but it used to work for years. It stopped working with my upgrade on my laptop to the latest version.

[BrianL]

IBM put a new "pull invalidation subsets" feature into 10.3 (cloud) and PA 2.0 that might be related.

You could try adding

Code: Select all

PullInvalidationSubsets=F

in your tm1s.cfg and see if that avoids the crash.

[Wim Gielis]

Bingo ! Thank you Brian, much appreciated.

[ndivine]

IBM put a new "pull invalidation subsets" feature into 10.3 (cloud) and PA 2.0 that might be related.

You could try adding

Code: Select all

PullInvalidationSubsets=F

in your tm1s.cfg and see if that avoids the crash.

Is this documented somewhere? Certainly isn't here: http://www.ibm.com/support/knowledgecen ... 503fe.html

[lotsaram]

For years and years, I use a TI process that:
1. Create a temporary dimension in the Prolog that and populate it (DimensionElementInsertDirect and the Update function)
2. Assign the ALL subset on that temporary dimension as the datasource of a TI process
3. Have some code in the Metadata tab, unrelated to the temporary dimension
4. Delete the temporary dimension again as the last line in the Epilog

This used to work fine. In the latest version of TM1 (Planning Analytics Local) the TI process crashes.
Without deleting the dimension in the Epilog, the process runs successfully.

Sounds like a regression bug. I can recall coming up against this in the past where either a temp dimension or temp cube was used as the data source for a process and if the data source for a TI process was deleted on the epilog of the TI then instant server crash. Sorry I can't recall from what version this bug was addressed.

Note *temp* object is used above in the traditional sense i.e. just a normal object that we delete when we no longer need it. In PA I would suggest changing your process to actually create the dimension as a temporary object then there is no need to delete it in the epilog at all.

[BrianL]

Is this documented somewhere? Certainly isn't here: http://www.ibm.com/support/knowledgecen ... 503fe.html

Unfortunately, I can't find any (official) documentation. I heard about it in a session at IBM's Vision conference. Not sure if there's still a link to download the presentation materials or if it was ever available publicly.

[Wim Gielis]

In fact, it crashes when I assign the subset 'ALL' and then delete the dimension in the Epilog.
If I create a different subset and assign it (even not temporary in the sense of the extra argument in the SubsetCreate function) then it does not crash and it works fine.
So that will be my change to the TI process.
Thanks everyone for replying.

[mvaspal]

Hi

In Planning Analytics on the Cloud, we had many crashes in various scenarios that have been fixed by:

Code: Select all

PullInvalidationSubsets=F

The server was quite unstable without that line, we had weeks when a crash happened ca. every second day

since then there is a new IF, called IF2, which solved these crashes without adding that cfg line (we had at least 4 different scenarios, neither of those were the scenario described in this thread)