dsmserv high memory usage

[Suslik]

Hi there,

since two weeks I have some trouble with the dsmserv process on our TSM Server V6.3.4. The server is an IBM P7 with 16 GB RAM and in the middle of the night the process takes the whole phy. memory and about 30 GB paging space. Last weak unix killed the process, because of not enough paging space. So we raised the paging space to 50 GB. Is there a simple way to say, push the phy. RAM, to solve the problem? How can I analyse, how many RAM I minimum needed for such an TSM?

Thanks for your help!

 

[moon-buddy]

Hi there,

since two weeks I have some trouble with the dsmserv process on our TSM Server V6.3.4. The server is an IBM P7 with 16 GB RAM and in the middle of the night the process takes the whole phy. memory and about 30 GB paging space. Last weak unix killed the process, because of not enough paging space. So we raised the paging space to 50 GB. Is there a simple way to say, push the phy. RAM, to solve the problem? How can I analyse, how many RAM I minimum needed for such an TSM?

Thanks for your help!

How many nodes are defined on this server? How many cores are on the P7?

In any regard, my TSM 6.3.3 servers runs on Redhat 5.8 on an Intel platform with 64 GB of RAM and 16 cores of Intel CPU running at 3.2 GHz. Load is ~425 nodes.

 

[Suslik]

How many nodes are defined on this server? How many cores are on the P7?

In any regard, my TSM 6.3.3 servers runs on Redhat 5.8 on an Intel platform with 64 GB of RAM and 16 cores of Intel CPU running at 3.2 GHz. Load is ~425 nodes.

My TSM 6.3.4 runs on AIX 6.1 with Power 7 CPU -> 4 Cores 3,6 GHz and 16 GB RAM and 800 nodes.

 

[moon-buddy]

My TSM 6.3.4 runs on AIX 6.1 with Power 7 CPU -> 4 Cores 3,6 GHz and 16 GB RAM and 800 nodes.

You may want to increase that memory to at least 48 GB. Cap the TSM usage to 80 percent of that via the dsmserv.opt file and restart TSM.

On an earlier job, with TSM 5.5.6, I have a P 520 with 4 cores and 24 GB of RAM serving just 325 nodes. I still saw memory peak usage from time-to-time but nothing disastrous.

 

[Suslik]

You may want to increase that memory to at least 48 GB. Cap the TSM usage to 80 percent of that via the dsmserv.opt file and restart TSM.

On an earlier job, with TSM 5.5.6, I have a P 520 with 4 cores and 24 GB of RAM serving just 325 nodes. I still saw memory peak usage from time-to-time but nothing disastrous.

Ok, thank you for your help. I discuss the server RAM upgrade in our meeting.