Cheshire Cat Computing

Hello,

Over time Perl.exe under processes in task manager on my server build up and run the box into the ground.

Is this a bug? Or, am I trying to do too much with this solution?

It seems like these processes should be cleaned up after they run but they are hanging around using up memory.

I am very inexperienced with this bundle but that's actually why I chose it. It's pretty simple to use. I've been able to build my graphing solution out to 36 devices so far and have more to add.

Today I rebooted the box with about 3500 perl.exe processes that were each using about 3k of memory. This was after about a week.

I really like this solution and have it setup pretty well. It's become a useful and invaluable tool for me and my dept, but sucks that every few days I need to reboot the box to clear those hanging processes out. I suppose I could set up a scheduled task to reboot the server every night. : )

Any help with this issue would make me as happy as an Network guy could be. Thanks!

This definitely sounds like a problem with your perl processes not exiting correctly.

Although MRTG can be a memory hog (ig you have a lot of devices, the multiple perl instances can eat memory like you wouldnt believe) it shouldnt be running so many instances.

You say 36 devices, but how many monitored targets (the real indication of load) do you have? Also, how much memory in the server? You should put in as much as you can - our MRTG server currently has 2Gb and we're upgrading it to 5Gb, although it deals with 4500 targets so has slightly higher requirements.

I dont run the Bundle myself, but hopefully someone else here who does can help with this.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Hi Steve,

Thanks for the quick reply.

If by targets you mean each interface I have about 800. The equipment I have is a Dell 850 with 2gigs of memory.

Aaron

This should be enough memory for this sort of load (note that I'm used to sizing Linux systems, windows requires more memory and CPU to run all the fluff). If you're running MRTG in daemon mode (I dont think you can run threaded under windows) then you will spread the load over the 5-min windows.

If you have set a 1-min interval rather than the default 5-min interval, though, you'll probably see the memory run out and the processes overspawn.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Sorry for my ignorance Steve. I believe it's at the default level. But in order to check what file would that configuration be located in? I did make some changes to the routers2.conf file.

Also, would the overspawning of processes cause them to hang around and not clean up? That's the core issue the box is having.

Thanks for your help!

Aaron

The interval is set in the various MRTG .cfg files. If you did not explicitly set it, then you've still got the default 5-min.

Since I would assume you have not defined custom plugins to monitor special things, the processes must be MRTG perls rather than plugins (if you dont know how to define a MRTG data collection plugin, then you dont have any).

Running out of memory would cause excessive swapping, it should not cause processes to overspawn.

Is it possible that you have the RunAsDaemon=yes set in the .cfg files, but are also calling the MRTG regularly with some sort of scheduler?

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11