Uptime Baby!

[der.hans]

Am 18. Oct, 2000 schwäzte Jason so:

> sinck@ugive.com wrote:
> > 
> > \_ If I had a business machine, I would try and have a cron job reboot
> > \_ the sucker on sunday night, at a random time between 3 and 3:30 AM
> > \_ when the load averages have been under .0? (whatever, depends on the
> > \_ CPU, etc etc) for at least 15 minutes. Why random?
> > Well, how about just "Why?"
> 
> If the machine is running unattended, it cleans out any processes that
> for one reason or another fail to terminate.
> 
> Netscape is perhaps the most heinous example of this (though that
> wouldnt be running unattended...) .. a crashed netscape can persist
> even after you exit from X!

There are better ways to hunt down spinning processes. A simple method is
to note that it's eating time. Prob is that Netscape also does this when
being used :(.

Doesn't it lose assocication with a tty when this happens?

sh-2.01$ uptime
  1:50am  up 259 days,  2:23, 34 users,  load average: 0.00, 0.01, 0.00
sh-2.01$ ps aux | tr : ' ' | sort +10 -n | tail -5
root       435  0.0  0.3  3004   816  p2 S    02 03  22 01 xterm 
lufthans   347  0.0  1.1  7572  2928  ?  S    01 59  23 44 kwm 
root      2683  0.0  0.2  2884   612  ?  S    01 50  90 31 xterm 
root     30178  1.2  0.1  1668   452  p3 S   Oct  7 182 47 ssh spliff 
root       176 99.9  9.6 36824 24988  ?  S    01 55 967 01 /usr/bin/X11/X

X has been up since I booted, so almost 300 days. It still hasn't broken
1000 minutes.

If it's a desktop you can probably trigger on anything hitting more than
100 minutes. If it's a server, maybe try 100, but be prepared to move to
500 or 1000 for the services the machine is providing.

If you've got spinning processes, check them by hand. You can probably
just kill Netscapes. If it's a service like nfs or apache you probably
want to investigate why it's spinning, not just ignore the prob.

ciao,

der.hans
-- 
#  der.hans@LuftHans.com   home.pages.de/~lufthans/   www.Opnix.com
#  You can't handle the source! - der.hans