Feeder Trouble

[BryanM]

Hoping for some enlightenment on a feeder issue.

I have three cubes, a main data cube (A) and two cubes (B&C) which extract some data from the main cube manipulate it and then it is pulled back into the main cube in a different place. B&C are almost identical in what they are doing, essentially just doing the calculation for a different subset of data. However, the data from Cube B is pulled successfully into Cube A and the values are fed. Data from Cube C is pulled into Cube A, but the values are not fed and so no consolidation is being performed.

Below are the pairs for rules and feeders, the first pair work and the second pair don't. As you can see the only real difference in the rules/feeders is the change territory from CON to DOM CON.

Rule from Cube A

['Adjustments','Local Currency','Actual','CON','43000 PI Recovery Variance - A-Series'] = N: DB('Variances Volumes','A Series','Net','Volume Variance',!Year,!Months);

Feeder from Cube B
['A Series','Net','Volume Variance']=>DB('Domino FC','Adjustments','Local Currency','Actual',!Year,!Months,'CON','43000 PI Recovery Variance - A-Series');

Rule from Cube A
['Adjustments','Local Currency','Actual','DOM CON','43000 PI Recovery Variance - A-Series'] = N: DB('Variance Volumes DOM','A Series','Net','Volume Variance',!Year,!Months);

Feeder from Cube C

['A Series','Net','Volume Variance']=>DB('Domino FC','Adjustments','Local Currency','Actual',!Year,!Months,'DOM CON','43000 PI Recovery Variance - A-Series');

skipcheck has not been turned on in cubes B&C so no feeders from Cube A to these cubes are required (I believe)



Version is TM1 9.1 SP2 U3

Thanks in advance

Bryan

[Steve Rowe]

Hi Byan,

So A is the central cube, you pull (rule) some values from A into cubes B and C do some work using rules and then pull (rule) the results back into cube A again?

If you don't need to view the results in B and C and the results that you pull back into cube A are not consolidations in cubes B and C then you don't need to skipcheck or feed cubes B and C.

The feeders you show are I assume in the rule sheet for Cube A and feed cubes B and C, I don't think you need these since B and C are not skipchecked and so do not need feeding.

If I summarise you rule like this

Cube A Answer = Ruled Stuff in Cube B or C = Starting Value in Cube A

Then you can see your feeder needs to be

Starting Value in Cube A =>Cube A Answer

and you can skip any feeding in cubes B and C all together.

HTH and I've understood what your trying to do.

Cheers,

[BryanM]

So A is the central cube, you pull (rule) some values from A into cubes B and C do some work using rules and then pull (rule) the results back into cube A again?

If you don't need to view the results in B and C and the results that you pull back into cube A are not consolidations in cubes B and C then you don't need to skipcheck or feed cubes B and C.

The feeders you show are I assume in the rule sheet for Cube A and feed cubes B and C, I don't think you need these since B and C are not skipchecked and so do not need feeding.

Hi Steve,

Thanks for the help. Yes, A is a central cube from which B & C pull some values. These values are used in some rules and then the results are pulled back into Cube A.

The feeders shown are in the rule files for cubes B & C respectively as cube A is pulling data from them and Cube A has skipcheck turned on. Cube A does not feed Cubes B or C.

I did think about feeding directly from Cube A as you suggest but didn't go down this route. The reason - Cube A and Cube B have been setup this way for while and working fine. I needed to replicate what Cube B is doing but with a different subset of data from Cube A. So, I set up Cube C and followed the same format for the rules and feeders.

Thanks,

Bryan

[Steve Rowe]

I'd be very nervous about feeding from cubes B and C without having B and C properly skipchecked and fed themsleves. Not sure how this would actually work.

If your pretty sure that everything in your new bit is the same then watch out for C levels that are involved in the new bit. Remember feeders only happen at the N level in the system and if you reference a C level in a cube it's just short hand for all the n levels under it.

Cheers,

[BryanM]

Feeder from Cube C

['Volume Variance']=>DB('Domino FC','Adjustments','Local Currency','Actual',!Year,!Months,'DOM CON','Blah');

I appear to have fixed the problem but not sure as to why it's decided to work now. I changed the feeder in Cube C as above to overfeed to 'Blah' in Cube A, a consolidation of codes that are affected. Checking Cube A after I made this change I could see that the consolidations were now working and values were fed.

Then I deleted the above line from the rule file and then unloaded the cube and loaded it again. Values in Cube A remain fed and the rule file is the same as it was this morning when they weren't.

Confused.

[Steve Rowe]

Feeders can get very confusing especially after you start unloading cubes and so on. Can I suggest you restart the server? It's really the only way you can be 100% sure of what the normal behaviour of the system is.

If you think of feeders as flags in cells rather than a live link like a rule then this may help you figure outwhat is going on / wrong.

You are not clear about which cube you un loaded but I from what you say I think it must have been cube C.

What's happening is this.

1.You over feed Cube A with your feeder in cube C, this populates the feeder flags in cube A and the calculations perform as expected.
2.You unload cube C, this has no affect on the feeder flags in cube A and you still get the results you expect.
3.You remove your over feed feeder from the rules of cube C and re load the cube. This refires the feeders from cube C to cube A but will not clear out the feeder flags already in cube A from step 1. You still get the results you expect.

If you restart the server you will find that cube A stops working because the feeders from step 1 have now been cleared from the system.
If you need to do this by unloading cubes, then you need to unload the cubes in the following order.

Unload cubes A, B and C.
Load cube A so it is in memory ready to recieve feeder flags when you load cubes B and C.

I'm still no nearer understanding why your original feeders are not working, it maybe because of the unskipchecked nature of cube C.

['A Series','Net','Volume Variance'] is this a C level of ruled values? If so I'd say that the feeders in cubes B and C do nothing and that the values from cube B that are working in cube A must be being feed from another location.

I'd probably just feed within cube A and simplify the whole thing, I'm not convinced that feeding from an unfed ruled value will work even if the cube is not skipchecked.

HTH

[Steve Vincent]

Has any part of your dimension hierarchy changed recently? something going in or out of a key consolidation may have contributed to this, i've found a similar thing in one of my models this morning. Also, does the data manipulation have to be done with a rule? If the data the calculations is based on isn't changing too regulary then maybe TI might be an option instead?

[BryanM]

Yes, I unloaded Cube C. I thought this would remove the feeder flags but it makes sense that it wouldn't affect them in other cubes. 'Volume Variance' in Cube C is a consolidation of ruled values.

To test this a bit further, I removed some of the feeders from the rule file for Cube B and left the rule file for Cube C as it was when I started this thread. Then I restarted the server. The values in Cube A that come from Cube B are still being fed and the values from Cube C are again not fed.

I've reverted to overfeeding from Cube C as per my earlier post while I try to see where the values in Cube B are being fed from.

The dimension hierarchy hasn't changed recently so I don't think this is a factor. The manipulation is done with a rule for ease of use - the data the calculations are based on changes each month as users submit their results.

Thanks for the help.