<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Stefan Gofferje wrote:
<blockquote cite="mid:48B7DD91.5080606@gofferje.homelinux.org"
type="cite">
<pre wrap="">Ivan Toman schrieb:
</pre>
<blockquote type="cite">
<pre wrap="">I don't think that using low priority nice will help, because if there
is enough free CPU resources, kernel will still give lot of CPU cycles
to your grads process. For example, think about it like this: You have
grads dedicated machine and don't use CPU for anything except base
system (relatively idle) and grads, but you run grads with nice=15. What
will then be CPU usage of grads? Theoretically it will top one of CPU
cores with near 100% utlisation, regardless of niceness=15. If you have
now some kind of watchguard on system that has trigger set, for example,
on 20% CPU utilisation over 3 seconds or more, it will kill your process
even if it has lowest priority. It all depends how watchguard is setup.
Try that cpu limiter that I posted link to, I think it can work for you.
</pre>
</blockquote>
<pre wrap=""><!---->
It is not about plain CPU-usage in %. During the night time when the
server is not very busy, the process goes through faster and is nerver
terminated. It's only terminated when GrADS is running AND some other
stuff also takes more ressources. I guess, it's either CPU _load_ or
memory but I wasn't able to find out from the admins.
</pre>
</blockquote>
<br>
I see. Do you have some kind of bash script that loops through grads
scripts? My idea is now to put some timeout (sleep x) in the end of
each loop step, so you may eventually avoid being killed by watchguard.
Or, if you have one big grads script for everything, you can put inside
it commands that will call external shell:<br>
'!sleep x'<br>
<br>
where x is time in seconds to wait.<br>
</body>
</html>