(ObjectNotFound) <cubename>: Error getting subset properties

[Lukas Meyer]

Hello,

This is the same problem as in this thread I recently started:

http://forums.olapforums.com/viewtopic.php?f=3&t=1391

so the systems settings remain the same: Server & Client 9.1 SP3, Perspectives with Excel 2003 SP2, English language in both, Client & Server

When a client double-clicks the cube to open the CubeViewer, a message box opens, reading "(ObjectNotFound) <cubename> : Error getting subset properties". You can click it away to catch a glimpse of the CubeViewer opening and closing immediately... You can find the screen-shot attached below, I censored the servername away since I don't know whether I am allowed to show it or not - better to take no risks
As you can see in the screen-shot, the first dimensions drop-down is empty, the first dimension is the company dimension which happens to have a perfectly fine working default subset, at least when opened in SubsetEditor.

In this case I was able to open some Views of this very cube - when I assigned the company-Default subset to title, the same error occured after recalculation. The next biggest subset, "Standard" (German for default, a remainder of old days when clients had a German TM1 version), would work perfectly fine.
My next step would have been to create a subset of the elements contained in Default but not in Standard, check if the subset produces the same error, and then delete elements from it until the new subset won't throw an error. Then I would have known which element is bad, but I don't know what I would have done with this information. Unfortunately some colleague solved the problem by recompiling the dimension via xdi before I could narrow down the error -.-

(saving the Default subset as itself or as A and then Default wouldn't help; there are no alias-attribute rules in elementattributes, tough there are s-element rules)

Sadly I am positive, that the error will be seen soon again ...

Did someone find the root of the problem and eradicate it?
When this error occurs the next time, do you think my approach would help me? What do I do if I find the bad element(s)?

Sorry for the new wall-o'-text and the new topic, but topics having 0 answers are twice as interesting as such having 5 or more, at least to me (And additionally, this is the proper one)

Thank you,
Lukas

[Herman Moller]

Lukas,

A very interesting problem. Just a thought do you have a default view that your cube opens? Try deleting the default view and and see if this cures the problem.

[Lukas Meyer]

Herman,

No, there is no explicit Default-View. If there was, at least I assume, deleting it would not have solved the problem - as mentioned I can take a working view (any view not having the Default-Subset of company dimension) and render the cubeviewer useless by selecting the company's Default-View (either title, rows or column will kill the thing)

The dimension is updated via xdi - forgot to mention that explicitly. The dimension was updated this way a couple of hours earlier, but right after the update it worked just fine. (this is a fact, the editor of the dimension checked the result)


The users were rather unhappy when I told them not to fix it but rather inform me when it happens next time. I think they should really enjoy their forced coffee-break ... but maybe they just like entering and playing with figures, and unlike me don't have to struggle with extremely slow mstsc-sessions, which make me grumpy two hours after I start to work...

Thank you,
Lukas

[Steve Rowe]

Lukas,
I get an error message like this in version 9.0 when I rebuild a dimension (almost always by TI) and have a cube viewer open that references the dimension.

In my situation though once I OK the message and recalculate the view again I get normal behaviour, I know this doesn't help much but it may give you a clue in tracking the issue.

My ideas would be
1. Ensure all cube viewers are closed when updating the dimension.
2. Stop using an xdi to update the dimension, I can easily imagine that this way of updating a dimension since it is much older then 9.1 and is more likely to be buggy than TI or directly in the dimension editor.

My only other thought is that there is some kind of bug around legacy German client and the default subset. I'm not exactly sure which situation you are in but have you tried playing around with combinations of Default and Standard being present in the subset list?

I'm assuming that the default subset is not empty at the end of the update process? This would definitely break everything. This is quite easy to do if you are working with a dimension that changes a lot.

HTH

[Lukas Meyer]

Steve,

Having the CubeViewer open and altering the dimensions would be a synchronization problem, since the client has a now outdated reference, thus the problem is not persistent. (Your problem was only for those users who had the CubeViewer open, and it would run fine when the client restarted, right?)

The one thing I hate more than xdi is xru (xdi have an actual advantage over other dimension-altering tools - they are readable and you can insert comments really easy - where xru just suck and should be wiped from existence, maybe even with fire or other drastic measures) - and I'm in no position to make the colleagues stop using them.

Yeah empty subsets, created by faulty processes, I create them sometimes, wonder why nothing works but usually find this mistake at step 2 of debugging.
So no, the Default subset does contain elements, as mentioned earlier it can be viewed with SubsetEditor. Copy elements & paste them in a new Defaulr subset did not work either, so the subset as such seems to be registered fine.

From earlier experience, the language-specific elements (hooray for no-alias-on-structural-items-like-applications-cubes-dimensions-views-and-subsets policy -.-) are protected and considered as essential only for the Client that accesses them (I can create an All-Subset in German Client, and I can create an Alle-Subset in English client, I can delete the created-statements section (which was written in English, but no worries - it got a doppelgänger, which, as a matter of fact, is very annoying if you don't get rid of it)). So it should do no harm for an English client to deal with a subset called Standard, although I can request to have the Standard subset renamed/deleted

I really hate having no clue at all...

Summarizing:
Error will stay after client restarts.
Error will be resolved when server is restarted.
XDI will be used to alter the dimension, even in future.
TI will alter the dimension. (The TI process did not fail between last XDI edit and error occurring)
I'll get rid of the Standard-subset (the German Default-subset)
I have not enough data to report this issue to IBM, they would probably laugh at me or ask me why recompiling the dimension is not good enough for me - I'm a snob indeed

thanks for your ideas, I'll amuse myself by starting a personal army of servers - and crash them