From: willmore@iastate.edu (David Willmore) Subject: Re: Intel, the Pentium and Linux Date: Thu, 29 Apr 1993 02:38:26 GMT
garrett@sba70.berkeley.edu (Garrett D'Amore) writes:
>willmore@iastate.edu (David Willmore) writes:
[... all *kinds* of stuff deleted ...]
>>I'm just looking around our network here for an example or two. Ok, here's
>>a ReadOnly NFS server that's been up since Oct 1 1992. There are several
>>more here with times in Nov and Dec '92. Anyone have a Linux system *that*
>>stable? If you do, you're running .95 or earlier and therefor having that
>>system up proves nothing about the stability of the current release of
>>Linux.
[... more things deleted ...]
>I think you have to realize a few things (referring to the original poster):
> 1) Linux isn't that old -- you can't expect that its going to work
> perfectly from its very inception. A little while ago I was unwilling
> to use it or 386BSD because they were "unstable". I am now running
> Linux quite happily.
I don't. My statement of the uptime of that server was to show that comercial
Unixes *are* stable, not so show that Linux isn't.
> 2) Unlike an older commercial Unix, there are hundreds of people working
> on various things for Linux, and posting their work (with SOURCE) on the
> net on a regular basis. Many people (myself among them) reboot simply to
> try out a new version of the kernel, or a different init. I've been in the
> habit of not rebooting my box unless I'm trying out a new kernel. And thus
> far I've not "needed" to upgrade the kernel. The only "necessary" upgrade
> I made so far was to upgrade my shared libs back when libc.4.3.2 came out.
> The majority of the Linux group are hackers, and we like futzing around with
> the system. Most of us wouldn't let even a "stable" OS go that long without
> upgrading device drivers, or inits, or doing something worthy of a reboot.
I fully understand this. I don't think Linux is any more unstable than a
system that is changing as rapidly as it is should be. The post that I
was responding to was the one that claimed that Linux was some kind of
stability god. It's not. Compared to comercial Unixes, it isn't. Compared
to other systems of comparable growth (BSD386) it is very stable.
> 3) Until Linux version "1.0" comes out, no one is "promising" a stable system.
> We are only reporting our own experiences. It is very likely that no one will
> ever "promise" a "stable" Linux system. Change is a fundamental part of
> improvement. If you have systems that haven't been restarted since Oct. 1992,
> probably you're administrators are not working very hard at doing what they're
> getting paid to do. I have a difficult time believing *any* UN*X system is so
> perfect that somebody isn't going to need or want to reboot, if just to add a
> device driver. I wonder what upgrades or fixes your sysadmins have missed?
> I do realize that in a production environment, the "if it ain't broke..."
> philiosophy is a driving force, but Linux is intended as a "research" vehicle,
> and that philosophy doesn't make sense for the Linux owner.
We have over 650 workstations and servers in our system. Not all of them are
effected by *every* bug. When a bug is fixed, the affected systems are updated
and brought back into working order. We don't go and bring down every machine
and install a new kernel just because some other machine was bothered by a bug.
We wait until each particular machine has to be brought down, normally for other
reasons, and do the update then.
Maybe you haven't ever run a production system before. Uptime on network wide
resources is important and you just don't go out and reboot 650+ machines because
you found a bug that only affected a certain class of machines. That's stupid.
Why would every machine need to be rebooted to add a device driver? Did that
device all of a sudden pop into existance on that machine? If it doesn't need
the driver, why take it offline to fix it?
If you would like to make any more personal attacks on our systems department,
feel free to try. They'er a very hard working lot and have much better things
to do than bother with new device drivers. Ever run a distributed computing
environment of 650+ machines? You might try it some time.
Later,
David
-- =========================================================================== willmore@iastate.edu | "Death before dishonor" | "Better dead than greek" | David Willmore | "Ever noticed how much they look like orchids? Lovely!" | ===========================================================================