Out Of Memory Error

[ssp]

Hi,
I was wondering if anyone has encountered the error where a Chore executes however at the completion of the chore an "out of memory" error appears. If anyone has encountered this before:
a) What are the causes?
b) How to rectify the issue?

Thanks in advance.

[Olivier]

Hello,


I would have a look at the cube logging settings.

Usually when you process a consequent amount of data, you might wanna switch your cube logging off at the beginning of your first process of the chore and switch it back on at the end o fteh last process epilog.

This will avoid a big memory consumption for the server.

If it doesn't resolve your issue, i would suggest that you give a bit more details about your chore and focus on the process that is running when the memory error occurs.

Hope this helps,


Olivier

[Alan Kirk]

I would have a look at the cube logging settings.

Usually when you process a consequent amount of data, you might wanna switch your cube logging off at the beginning of your first process of the chore and switch it back on at the end o fteh last process epilog.

This will avoid a big memory consumption for the server.

I'm not saying that you're wrong here, but I have to admit that I've never seen logging affect the amount of memory used. The size of the log file, yes. The speed of processing (since writing to a file is a relatively slow process compared to in-memory operations), most certainly.

But the memory? I'm not clear on why that would be affected significantly.

In relation to the original question, the only thing I can think of is that the chore really is loading up so much data that the server runs out of memory. I'd suggest starting the performance monitor on the server immediately before running the chore. In this case you'd need to have logging on for the four }Stats cubes, since if the server's crashing out of memory (which is, I presume, what you're saying is happening) then taking a look at the logs is the only way you'll be able to see the values for those cubes while the chore is running.

(Normally if you have the performance monitor running you need to DISABLE logging on the four }Stats cubes; otherwise you end up with gigabytes of log files very, very quickly.)

The }StatsForServer and }StatsByCube cubes would be the ones of primary interest in this case.

[Olivier]

I do take your point Alan,
I got confused between several issues which were mixed in a past customer model...
I think you are right concerning the main impacts of the Logging.

Ssp did you manage to modify to chore to have it working properly ?

Olivier

[Paul Coggan]

Hi SSP

It definately sounds like your server (or your PC if you are running a local server) is running out of memory. I find that the fastest way to verify this is to open Task Manager on the server while the problem process is running. This will be more informative than the stats cubes because you need to consider the memory used by other processes as well as that used by TM1. To see the memory used by TM1, open the processes tab in Task manager and sort by memory usage - TM1 then will be at the top of the list. You may also need to select the box in the lower left corner to show processes from all users if the process is run under another user ID. The total physical memory used by all processes can be viewed on the Performance tab.

If you are running 32 bit version of TM1 you will be able to have up to 2 gig of memory allocated to TM1. (3 gig is possible if you have 4gig of memory on board and use a switch to make the extra gig available - see the old TM1 forum for this). If the total memory used by TM1 reaches 2 gig, or if you have less memory installed, the total memory used reaches the total amount installed on your server, then TM1 will crash. It will also slow down signifcantly once you get within about 20% of the limit.

If you are running out of memory you should look at whether your process is causing a huge jump in the memory used, or if you were already close to the limit before the process was run and it just pushed you over the top. Also take a look at reordering dimensions within your larger cubes as this can significantly reduce the amount of memory used by a cube. Then check the efficiency of your feedeers in any rules (you will need to restart the TM1 service to see feeder improvements take effect on memory). Otherwise you may need to buy some more memory. There are some other options if these dont work - let us know how you go.

These memory limits wont apply in 64 bit version of TM1.