Strange feeder behaviour

[vladkon]

Hello All, I am experiencing some strange behaviour with one of the feeders. There are persistent feeders enabled in the model. I changed the rules so that there are no feeders in this cube, stopped TM1 service and deleted .feeder file. I expected that there will be no .feeder file created after the restart, yet there still is. Also checked that there are no feeeders in other cubes pointing to this cube. Any ideas what else can it be?
TM1 10.2.2 FP7

[tomok]

Do your rules pull in values from other cubes? Perhaps you still have feeders in those cubes for the intercube references?

[PavoGa]

The feeder file will be created for a cube if any cell in that cube is fed. As tomok indicates, if another cube's rules are feeding the cube in question, you will get a feeder file as long as PersistentFeeders=T in the tm1s.cfg file.

[vladkon]

The feeder file will be created for a cube if any cell in that cube is fed. As tomok indicates, if another cube's rules are feeding the cube in question, you will get a feeder file as long as PersistentFeeders=T in the tm1s.cfg file.

That is what I thought, but I deleted all other rules files from the model. The only rule file left is for the cube in question and it has no feeders section!
The only other suspect is function ConsolidateChildren - could it be that it causes feeders creation?

[lotsaram]

The mere presence of a rule file, even if empty, will cause creation of a .feeders file for the same cube.

[vladkon]

The mere presence of a rule file, even if empty, will cause creation of a .feeders file for the same cube.

Interesting. Than what is stored in that file if not feeders info?
I am trying to recreate the model without feeders for currency calculations, to see if it will improve performance. Not sure what is the point of not using feeders, if there is still a feeders file created, even if there are no feeders...

[macsir]

Try to turn of persistent feeders in cfg file and see.

[lotsaram]

Interesting. Than what is stored in that file if not feeders info?
I am trying to recreate the model without feeders for currency calculations, to see if it will improve performance. Not sure what is the point of not using feeders, if there is still a feeders file created, even if there are no feeders...

Any cube with a rule associated with it must have a feeders file if using persistent feeders. This is necessary so that on session start the server can validate that the feeder file is same or later time stamped than the cube file. If the feeder file is older this will invalidate all feeder files and cause reprocessing of feeders. So even if a feeder file is completely empty it still needs to serve this purpose.

[vladkon]

Any cube with a rule associated with it must have a feeders file if using persistent feeders. This is necessary so that on session start the server can validate that the feeder file is same or later time stamped than the cube file. If the feeder file is older this will invalidate all feeder files and cause reprocessing of feeders. So even if a feeder file is completely empty it still needs to serve this purpose.

Understood. But it is far from being empty - it takes about 16Gb of disk space. Feeders as per definition should include only data on n-level cells, correct? If there is no feeder section in the rule file, what could it be that takes such an amount of disk space, and how it can be reduced?

[vladkon]

What I am trying to say, is that it seems not only feeder section controls feeders creation, but also rules some rules statements, for example ConsolidateChildren. If this is true, than it also should be taken into account when optimizing feeders.

[lotsaram]

What I am trying to say, is that it seems not only feeder section controls feeders creation, but also rules some rules statements, for example ConsolidateChildren. If this is true, than it also should be taken into account when optimizing feeders.

No. Only feeder statements create feeders. Rules and feeders are independent. Feeders are only used for consolidation, period. That's all they do.

If you have a 16 gb feeder file and the rule for that cube contains no feeders then (assuming the server had been restarted with feeders having been cleared with delete all persistent feeders command) the cube is being fed from another cube. Simple as that.

[rpar]

ConsolidateChildren function does create feeders. With no feeder statements if you're seeing feeders in the }StatsByCube cube, ConsolidateChildren is the culprit.