TM1 Forum

I have an issue with the (unhelpful) error message in the title. Hopefully some IBM lurkers here can enlighten as to 1/ what the error message actually means and 2/ the exact cause and suggested resolution
Error with spreading, error type unknown = 188

I am developing a model with a large data volume and so to speed up data processing throughput with cube to cube transfers I am using parallel interaction with TM1RunTI to split into multiple threads and run at the same time. Some of the processes use CellPutProportionalSpread not just plain vanilla CellPutN due to dimension asymmetry for some data transfers. The TM1RunTI threads run under a separate system "data loader" user that has read access to all processes but no access to cubes and dimensions. The problem is that when running under the data loader user the error message come up and data is not changed. It is a problem with security as opposed to parallelization as running the same process with the data loader user with a normal client session produces the same error message and using a full admin account and TM1RunTI is also successful.

It can be worked around by making the data loader user DataAdmin and I hypothesise that it is something to do with the failure of }Hold cubes being created as there is no Write access for the data loader user to the cube where the CellPutProportionalSpread is directed. However this is not optimal as from a security and compliance perspective the data loader user should only have access to the processes that it needs to run and no other objects on the server. In my view this is a bug but I would be really interested in what others think and whether from an IBM perspective this is a bug or "expected behaviour".

Wim pointed to a similar concept here http://www.tm1forum.com/viewtopic.php?f=3&t=10466 where it looks like the whole idea (documented) of TIs running with Data Admin access is actually a load of rubbish. I never actually got round to logging it with IBM though.

lotsaram wrote:I would be really interested in what others think and whether from an IBM perspective this is a bug or "expected behaviour".

Any rules? To prove the bug you'd at least have to show the error cropping up where the cube that was the target of the spread had no rules, even STET. Also, did you experiment with different settings for SpreadingPrecision?

lotsaram wrote:I hypothesise that it is something to do with the failure of }Hold cubes being created....

There is something funky going on in control-cube-land with holds as the }Hold cubes seem to pop up for people who categorically deny ever having performed spreads...

lotsaram wrote:the data loader user should only have access to the processes that it needs to run and no other objects on the server.

If this aspect is more important than using CellPutProportionalSpread then you can work around the need for object security by writing the TI to re-implement the proportional spread just with CellPutN's. This still isn't a great outcome

Yeah lots of rules. For the moment (and probably permanently) I have solved this by elevating the security profile of the system data loader user to have write access to the cubes the CellPutProportionalSpread is being run on. (The idea of creating some kind of allocation matrix to cycle through and using only CellPutN was not at all attractive).

I would still be interested in hearing what exactly error type unknown = 188 means. For that matter also error type 185 which also crops up.