26 Jan
2010
26 Jan
'10
12:17 a.m.
On Jan 26, 2010, at 2:56 AM, Alexander Schreiber wrote:
For legal reasons: no. And if that didn't get in the way of it, not
without the express permission of the owner, for not wanting to act
like
a digital hooligan.
Ahh, my faith in you is restored. ;) But I did just upgrade sshd
on that machine just in case! =)
Heh, unfortunately! ;)
But what I was really referring to is operating systems needing to
be preemptively rebooted to keep them stable, and other absurd stuff
like that. Nobody should put up with that. The fact that many
people DO put up with it doesn't change the fact that it's just plain
silly.
-Dave
--
Dave McGuire
Port Charlotte, FL
IP address please[0], there probably were a few remote
kernel
bugs in
the last 3 years.
Do you really want to make a run at it? Well, usually the thing that gets in the way of high
uptimes these
days
in proper systems is not the machine crashing, but either upgrades or
security patches, including ones were a reboot is needed to run a new
kernel.
The latter situation is the only case where I end up doing it.
With modern kernels being so ridiculously module-oriented, I almost
never really need to reboot, though.
Another thing killing uptime is the bloody hardware
failing.
I generally don't experience that.
Modern hardware, and especially disks, still fail too
damn often. Have
enough machines up and running and you'll lose a disk almost every
day.
And that is with the good (high MTBF) stuff.
Yes. Years ago, I managed a network of over 1,000 machines
(mostly Sun SPARCstation-2s with some -5s sprinkled about) and I did
see some disk failures. They involved downtime, a screwdriver, the
spinning up of Exabyte tapes, and a lot of cursing in the
datacenter. Very annoying. Nowadays, thank heaven, I just replace
the failed drive and watch the array resync while sipping a latte and
reading an issue of EDN, and the customer never knows it happened.
I've done rolling upgrades of system disks (as in, completely
replaced boot volumes) without shutting machines down that way. I
rotate drives out of service *usually* before they fail, but when
they do, there's no downtime involved. I very, very rarely reboot
most of the computers that I'm responsible for because, most of the
time, it just isn't required.
Amazingly, I
also expect my car to not randomly stall or catch
on fire
while I'm driving down the road.
Different rules apply there. Cars randomly catching on fire tends
to be
expensive for the manufacturer, disks dying after two years doesn't,
usually.