At 9:45 AM +0100 9/3/04, Andy Holyer wrote:
I explained that generally some upgrade comes along that requires
a reboot, but I realized that I don't know how long a box would
stay up in the maximum. So, come on, this should be fun, what's
the biggest uptime you've ever had for a BSD box?
I don't think it would ever "require" a reboot. The question is
whether you need to reboot to apply some prudent updates and
I have one server that I try to keep up as much as possible. The
three longest runs on that machine are:
373 days 10 hours, ending in July 2000 (long power outage)
599 days 14 hours, ending in Sept 2002 (UPS failure)
497 days 18 hours, ending in Apr 2004 (disk failure)
The first one ended because a power-station going into campus was
flooded (due to some construction in the area), and the building
did not have any power for about four hours. My UPS lasted about
three and a half hours before giving out.
The second one was that the UPS itself melted down! Well, it did
not quite melt, but it was seriously overheating and I had to
shutdown all the machines connected to it and unplug everything.
The UPS was literally too hot for me to touch, and once it cooled
down enough (which took about four hours), I could see that the
battery had started to melt.
The third was a disk problem, but I also believe it was a OS error
because the disk *getting* the error was one I should have been
able to ignore. However the OS was confused over which disk got
the error, and it kept resetting the disk-controller for the main
system disk, instead of the one for the disk which had the errors.
So, I suspect the fault for that reboot is half hardware and half
the OS itself.
If you are going for long up times, then the stupidest thing you
can do is "install it and forget it". While I have long uptimes
on this machine, I also have only a few network services running,
and there are only two or three people who can log onto the
machine (and I trust them). I use the ports collection to keep
many things up-to-date, and for some things in the base system
(like sendmail), I recompile them on a different machine and
then copy the pieces over to this server. So, I manage to apply
the vast majority of security fixes, even though I do not reboot
and I do not have to stop/restart the main service that this
Garance Alistair Drosehn = [EMAIL PROTECTED]
Senior Systems Programmer or [EMAIL PROTECTED]
Rensselaer Polytechnic Institute or [EMAIL PROTECTED]
[EMAIL PROTECTED] mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"