I am lost as to why my Red Hat 9, Dell 6650, Quad 2.4 GHZ Xeon, 16GB RAM server with a fiber direct attach EMC Clarion array is eating into over 1GB of of a 2GB swap partition. Common sense says that you put 16GB of RAM in a server and swap should never be needed.

Anyone have experience/input regarding why this is or what I might be doing wrong?

Just a guess, but maybe your kernel can't handle all of that ram. Like i said, just a guess.

issue the command swapoff to your swap partition and see if it causes any problems after a while.

You don't mention this, but you did recompile your kernel to tell linux that it has more mem then the average pc??

Processor type and features -> High Memory Support -> 64GB

There are some other option, check them out too.

Hi,

Usually there is more than 70-80 services running on the Linux OS...they are basic and system services actually all of the services...but this services are not all the time up...they are in status "sleep"...so they are in the swap... And sometimes the RAM is not the sure est place for collecting info so the intelligent Linux decide to keep the temporary info into the swap...If you make the swap 500Mb or smaller there won't be problem but Linux will calculate it and make the system again optimized...

Gorbachov

I am running Red Hat 9 kernel 2.4.20-20.9bigmem so the kernel should be able to handle the 16GB of RAM. When the system is on its own, doing day end reports and fully utilizing all processors, the swap is hardly touched, but when we have about 250 telnet sessions accessing our Progress database, the swap grows substantially. We are also experiencing slowness. On our previous Red Hat Advanced Server 2.1 kernel, the swap barely went over 50MB but the server hanged often.

I know that one of the few kernal varients you can set with the RH installation disks are a large memory one called BigMem. You may want to enable this for over 2GB, you don't need to develop a custom kernal for it.

Also have a look at the output of top to see what's eating all your memory.

Before I spill my 2 cents: 250 telnet sessions?? Must give you a very save and secure feeling

I've read, more then once, that some versions of telnetd have bugs (memory leaks among others).

From the man page:

Because of bugs in the original 4.2 BSD telnet(1), in.telnetd performs
some dubious protocol exchanges to try to discover if the remote client
is, in fact, a 4.2 BSD telnet(1).

Another good reason to get rid of telnet..........

Drunna:
Okay, I will admit I am a newbie on Linux(less than 1 year), so don't laugh when I ask what other option do I have other than telnet that will allow dummy terminals hooked up via serial connection to access my Linux system? I use SSH to do my work from my laptop.

Gorbachov:
I see on my top output that I have about 1450 processes with 1400 of those sleeping. Is this what you mean? Linux is putting those 1400 sleeping processes in swap?

Lastly, we are running a product called Unform that allows us to print to laser printers. We are getting significant zombies (50) that seem to be related (per the PS command). Could this be related to the high swap useage?