Major upgrade failure -> people maintain > 1 computer

[Marko Vojinovic]

On Wednesday 07 December 2011 23:48:58 Reindl Harald wrote:
> Am 07.12.2011 16:50, schrieb Marko Vojinovic:
> > On Wednesday 07 December 2011 14:39:45 Reindl Harald wrote:
> >> did you ever work in an environment with a lot of servers and
> >> users and used rsync / nfs?
> > 
> > Why would you even consider using Fedora in such an environment? If you
> > have a server farm with shared users and use rsync/nfs/whatever, and
> > you have the whole thing (or a part of it) running on Fedora, then
> > you'd better be prepared to do some nontrivial amount of work when
> > upgrading the Fedora machines.
> 
> becuase it works perfectly if you know hat you are doing

Sure, if you know what you are doing, you are welcome to do it. But it may 
just happen that future Fedora releases prove you wrong later on. Fedora is a 
fast-moving target, and you can never be sure in which direction it is going 
to go in a year or more.

> because using stoneold software like RHEL is not a option here

What, you are missing the latest Gnome3.2 on your production servers? :-)

More seriously, I agree that sometimes there is a legitimate need for the 
latest software in production, but those cases are exceptional rather than 
typical.
 
> > And yes, that's exactly what I mean --- *work* --- create and execute a
> > script to chown across all disks on all machines, update/modify all
> > /etc/passwd and /etc/group to reflect the UID+500 change on all systems
> 
> for exactly what reason?
> because anybody thought it has to be changed?

And what do you think, *why* did Fedora decide to raise the limit on UID's in 
this release? Is it not reasonable to assume that they have a plan to actually 
*use* more than 500 system-reserved UID's in some future release? My guess is 
that in a couple more Fedora releases there will be *no* *choice* but to raise 
the UID limit. Consequently, hoping that you can get away with current UID's 
in the long term, while upgrading through Fedora releases, is not very wise 
IMHO.

> > It will require some downtime
> 
> what is this? downtime?
> 
> this weekend i did a upgrade from F14 to F15 on all our servers
> and the downtime was exactly 30 seconds for the reboot and the
> upgrade per machine takes between 5 and 7 minutes (maximum stripped
> down installations on each server and good hardware)

It's great if you can do that. But there are also other usecases, where 
serious downtime may be necessary.

> > rebooting of all servers (maybe  simultaneously)
> 
> hopefully you are not responsible for production environment

Hopefully you are aware that Fedora 16 has gone through at least four kernel 
versions already, since it first came out (might be even more, I didn't count 
too carefully).

And please don't tell me that you don't need to run the latest kernel on 
production systems where the latest software is an absolute must-have. I know 
that in some cases you can get away with that, but again it's an exceptional 
rather than a typical usecase.

Best, :-)
Marko