REPOS-E-NOMEM, Insufficient memory on machine for operation.

[ornel]

Hi,

I would like to generate one of my cube, but I got an error message.
Does anyone have any suggestions?
Here is the log file:

====================================================
PowerPlay Transformer(7.0.713.0) Thu Aug 07 14:57:02 2003

LogFileDirectory=e:\cognoseles\_automata
ModelSaveDirectory=d:\programs\Cognos\cer2\binDataSourceDirectory=d:\programs\Cognos\cer2\binCubeSaveDirectory=e:\cognoseles\kozponti\olapkockakDataWorkDirectory=d:\temp\;e:\tempModelWorkDirectory=e:\tempMaxTransactionNum=1000000
ReadCacheSize=32768
WriteCacheSize=32768
Temporary UDA directory=C:\temp

2003.08.07 14:57:09 4 0000016D Start cube update.
2003.08.07 14:57:09 4 0000016D Initializing categories.
2003.08.07 14:57:09 4 0000016D Timing, INITIALIZING CATEGORIES,00:00:00
2003.08.07 14:57:09 4 0000016D Start processing data source 'c:\teszt\customer_cube_sas.iqd'.
2003.08.07 14:57:09 4 0000016D Reading source data.
2003.08.07 14:57:11 4 0000016D Timing, OPEN DATA SOURCE,00:00:02
2003.08.07 15:12:09 4 0000016D 850000 records have been read; 823317 categories have been created.
2003.08.07 15:27:09 4 0000016D 1163000 records have been read; 1114340 categories have been created.
2003.08.07 15:42:09 4 0000016D 1639000 records have been read; 1556940 categories have been created.
2003.08.07 16:02:20 4 0000016D 1690000 records have been read; 1601301 categories have been created.
2003.08.07 16:17:22 4 0000016D 1735000 records have been read; 1641152 categories have been created.
2003.08.07 16:32:22 4 0000016D 1797000 records have been read; 1699152 categories have been created.
2003.08.07 16:47:23 4 0000016D 2270000 records have been read; 2127220 categories have been created.
2003.08.07 17:02:24 4 0000016D 2681000 records have been read; 2451523 categories have been created.
2003.08.07 17:17:44 4 0000016D 2748000 records have been read; 2492827 categories have been created.
2003.08.07 17:33:29 4 0000016D 2998000 records have been read; 2646196 categories have been created.
2003.08.07 17:48:33 4 0000016D 3141000 records have been read; 2776366 categories have been created.
2003.08.07 18:03:35 4 0000016D 3381000 records have been read; 3012442 categories have been created.
2003.08.07 18:18:51 4 0000016D 3475000 records have been read; 3100301 categories have been created.
2003.08.07 18:34:54 4 0000016D 3490000 records have been read; 3115544 categories have been created.
2003.08.07 18:50:28 4 0000016D 3505000 records have been read; 3129956 categories have been created.
2003.08.07 19:05:29 4 0000016D 3863000 records have been read; 3466958 categories have been created.
2003.08.07 19:20:49 4 0000016D 3989000 records have been read; 3579320 categories have been created.
2003.08.07 19:36:25 4 0000016D 4024000 records have been read; 3612189 categories have been created.
2003.08.07 19:51:33 4 0000016D 4070000 records have been read; 3654028 categories have been created.
2003.08.07 20:07:13 4 0000016D 4098000 records have been read; 3680083 categories have been created.
2003.08.07 20:22:35 4 0000016D 4120000 records have been read; 3700746 categories have been created.
2003.08.07 20:37:45 4 0000016D 4151000 records have been read; 3729375 categories have been created.
2003.08.08 06:03:47 1 0000016D REPOS-E-NOMEM, Insufficient memory on machine for operation. [->OK]


thx
ornel

 

[BruceReed]

ornel

I am sorry if some of these steps seem basic, but I am not sure what you have already tried.

This looks like a very large cube you are generating. Each stage is over 1 million entries (many are over 3 million), so the end cube will have about 85 million entries, which is rather large. Is this the first time you are building this cube? Is it possible that you are getting outer joins (cross product) in Impromptu? Have you checked out the source reports (I assume Impromptu) to see if they are all OK?

Have you checked out the basics, is your hard drive large enough, have you defragged it, have you cleaned out the temp directories, etc?

If this does not help, I would suggest that you summarize the data. What are all of those iqds? Do you need every detail of everyone of them? Can you live with out some of the details? You can use transformer to summarize old details, but keep the details for the newer data.

Good luck
Bruce

 

[BruceReed]

Duh

I added up the log entries to total to 85 million entries. However the log entries are cumulative, so my quoted number is complete BS. Sorry about that.

However my comments still seem OK.

Thx
BR

 

[ornel]

Hi, BruceReed

I have got only 1 table (no cross product query,
It's very large table, it has got lot of columns, and 6 million rows.
My Transformer model also very complex.
I tried to set WriteCacheSize (=65536), and I have got 102 GB free disk space, but I can not generate my cube...
I got the "REPOS-E-NOMEM, Insufficient memory on machine for operation." error message again and agian.

My server is Win2000 (dual pentium processor - 2.3 GHz, and 2 MB RAM).

Any answer is appreciated.

thx ornel

 

[BruceReed]

You certainly have enough RAM & disk memory, but you are running a very large report (over 3.7 million categories for 4.1 entries when it crashed). Do you need that amount of granularity in the cube? Normally cubes are used to condense the material down, and not have so many to worry about. PP is used as a 40,000 foot overview of your data.

Did you try partitioning the cubes or sub-cubes? This might help, but none of my cubes were ever this large.

Other than that, I don't have any other advice for you. Try Cognos directly for more help.

Sorry
Bruce

 

[BruceReed]

One more thing, did you try to run "test-build" in transformer to see if the end cube is what you expected?

I am still worried that you are getting too many categories for the amount of data you have. Perhaps a lot of the items are orphans, not in any category.

If I am barking up the wrong tree, let me know.

Cheers
BR