Fedora Makes a Terrible Server?

[David G. Mackay]

On Wed, 2008-03-26 at 08:36 +0900, John Summerfield wrote:
> Not Fedora exactly (it's the f9a), but yum has conspired with rpm 
> against me to remove every working kernel. yum ignored how many kernels 
> I wanted to keep, and (apparently) upgraded (as opposed to installed) 
> the kernel.
> 
> It _could_ happen in Fedora, to anyone who had just one kernel (eg a new 
> install).

Yes, and that's why I use a VM to install and test the new releases.  If
it completely fries the VM, I'm only out a few minutes time.

> > Having a VM running a server in this environment may not be the absolute
> > best practice, but it's certainly feasible.
> 
> That's not the point I was speaking about. Running a VM _under Fedora_ 
> is. A VM under CentOS or RHEL is altogether different. RHEL/CentOS is 
> less likely to break with a new update, and since it presents 
> conservative "hardware" to the guest, Fedora is less likely to break too.

What can I say?  I do have multiple machines, and have dedicated
servers, and I want something with a long lifecycle for those servers
doing fairly standard things, like dns, http, imap, samba, etc.  At the
same time, I like to keep tabs on fairly state-of-the-art developments
so that I can get a good idea about where my own development efforts
should go.  I have the luxury of multiple machines.  If I didn't,
running a VM with Fedora on the desktop and Centos as the guest would be
a likely choice.  I have some hardware (a tv capture card, for one) that
isn't implemented in virtualized hardware.  FWIW, I've had pretty good
success running QEMU-KVM out of rawhide with multiple guest OSes.  If I
relied on the version available in RHEL or Centos, I wouldn't be nearly
as happy.  In the case of QEMU/KVM, the bleeding edge stuff is working
better.  Of course, that's just my experience, and I guess that it's
possible that I've been fantastically lucky, but I don't think so.

Dave