limiting cpu usage time for daemons

[venkman]

I was wondering how to go about limiting the % of CPU used by certain processes. I've been running hotway, the pop-httpmail gateway daemon, as an xinetd service. Unfortunately, after a couple of uses, hotway starts spawning zombie threads which eat away at my cpu. I've been using cron to regularly kill the extra hotwayd processes, but I was wondering if there was a way to limit the damage done in the interim between the cron job.

I was thinking ulimit might be the way to do it, but I think that only covers user threads.

any ideas?

-Venkman

 

[un1xh3ad]

Hello

you might want to use the command nice. Man nice for full instruction , but heres an example:

nice -n 20 /path/some/program

will start the program with the lowest possible priority.

nice -n -19 /path/some/program

will start the program with the highest possible priority.

Cheers

Neil

 

[venkman]

but, it's not a process that I spawn from the command line. The process is started from xinetd. Can nice be worked into that without effecting other processs xinetd spawns... and if so how?

-Venkman

 

[venkman]

oh, and can nice be used to changed the priority of an existing process?

-Venkman

 

[un1xh3ad]

I am not sure it can be used for xinetd processes , i have not tried, but the file in xinetd names the binary , so try prefixing it. Or running them as daemons , not from xinetd , and then use nice.

Yes , a running program can be re-niced.

Cheers

Neil

 

[venkman]

oh, and I just realized that in my original post I wrote "zombie threads" when I really meant "zombie processes".

Thanks for the advice Neil. I'll try it out when I get a chance and let you know how it goes. I think renaming the binary called in the xinetd sounds like the best idea to me.

Also, admitedly, I have not checked the hotway site to see what they say. Probably should've checked that out first, but I figured this was a useful thing to learn anyways. If I see anything on the hotway site of interest I'll post it here.

Thanks again,
-Venkman