Re: PROPOSAL: Upgrading every RHEL server to RHEL5 on Sun April 1st

[Ross Golder <ross golder org>]

Jeff Waugh wrote:
> <quote who="Olav Vitters">
>
> > I propose to upgrade every RHEL server from RHEL3/4 to RHEL5 on Sunday
> > April 1 2007. Very likely things will break, that is why I NEED additional
> > sysadmins to standby.
>
> So, to start with: Thanks very much again for proposing this, and devising a
> step by step process to complete it.
>
> I am concerned, however, that while this process gets us from A to B (RHELx
> to RHEL5), it doesn't achieve anything regarding our problematic process and
> management issues.
>
> An untested, long slog to upgrade a whole bunch of stuff (even if it's not
> "everything") is going to be *hard*. You've laid out a good process based on
> the functional dependencies, but we don't even have an installation standard
> (which Matthew asked about) at this point, a few days before we start.
>
> I would feel more comfortable taking on one machine to upgrade, testing the
> services locally beforehand, then doing a very fast reinstall/reconfigure
> process to bring it back up. In this case, label is the simplest and best
> choice. Once it is running RHEL5, we can use virtualisation to assist with
> the rest of the process.
>
> The advantage to taking this more slowly is that we can institute standards,
> testing and documentation along the way, things that will MASSIVELY improve
> our efficiency and service delivery in the future.
>
> I'm committed to participating in the process, but I'd really like to make
> sure it's a process that is safe, sane, *and* can put us on a better course
> to the one we've been on for some time.
>
> If everyone decides we should run with Olav's process as-is, straight away,
> then I will commit to a second migration process, focusing on establishing
> what I've described above.

FWIW, I'm for running with Olav's process now, as he's evidently put a 
lot of thought into it and is willing to follow it through - I'll gladly 
do what I can to help ensure it stays on track whilst it's going on and 
follow up/track any loose ends that come out of it. We'll end up better 
off for it, for the sake of a couple of days of teamwork on IRC.

But I do also see what you mean about taking the slow approach for 
better testing and documentation procedures etc. I just feel that, at 
this stage, getting everybody on the team together on IRC for a day or 
two is going to be hard enough - if the process was any more involved 
and ended up taking up several weekends then less people on the team 
will be able to commit the extra time required.

So, IMHO, let's get our RHEL3/4 platforms up to RHEL5 now as we have the 
momentum, the opportunity and a plan, then we'll have plenty of time to 
prepare a more involved/improved upgrade process (maybe including 
virtualisation or whatever etc) which we can discuss and apply when it's 
time to do this for RHEL6, 7 and 8 :)

Cheers,

--
Ross