[WBEL-users] Upgrading from Whitebox to RHEL

[Robert Heller]

At Fri, 27 Apr 2007 10:54:07 -0700 Scott Silva <ssilva at sgvwater.com> wrote:

> 
> Mike B. spake the following on 4/26/2007 7:37 PM:
> > At 4/26/2007 01:09 PM, Scott Silva wrote:
> > 
> >> wasting your money. They will probably recommend a clean install and
> >> careful
> >> data restore.
> > 
> > This raises a different question...one that may be obvious to long-time
> > unix/linux folks, but not to us relatively new ones:
> > 
> > What is the best way to deal with upgrades and re-installations of
> > linux?  That is, how do you preserve all the configuration changes,
> > startup customizations, user accounts, cron scripts, DNS zone files, and
> > other system management "stuff" that accumulates on a system over time
> > when you re-install the OS or upgrade or switch to a different
> > distribution?  Manually re-customizing and repeating all the setup work
> > may be possible, but it's not an attractive option at all.

The main thing is proper partitioning.  For *my* thoughts on 'proper
partitioning' see

http://www.deepsoft.com/Articles/

and select the 'Partitioning for Linux' article.

Almost all of the 'system management "stuff"' lives under /etc, with a
small handfull under various places under /var somewhere (eg /var/named
for DNS zone files, etc.).  If stuff like /home is on its own partition
and if /var/spool, /var/named, /var/www and so on are on their own
partitions as well, you only really need to make a backup of /etc some
where -- I usually create a /scratch partition and when I need to do a
fresh install / big leap upgrade (eg from RH 7.3 to CentOS 4.3, which I
did last year) / change distro, I backup /etc/ to /scratch/etc/... 
Making backups of things like /usr/local or really the sources to the
stuff installed there, since more often than not the fresh install will
involve different libraries and things, which will require rebuilding
(*I* just keep these sorts of sources on some other partition (like
/home or /sources or something like that).

> > 
> > When you upgrade within a distro this is usually not an issue...files
> > don't get deleted, upgrade scripts usually rename them if they get
> > replaced with upgraded ones so you can merge in whatever is still
> > relevant, and user directories aren't touched, etc..  When you do a
> > clean install, this isn't the case.
> > 
> > Getting listings of files changed since you installed before you shut
> > down the old system might be one way to catch what you need to save, but
> > it will also catch a lot of stuff you may not care about (old logs,
> > backup copies of edited files, programs that are going to be replaced
> > anyway, programs you don't want anymore, etc.) and weeding it isn't much
> > more attractive than repeating all the work.
> > 
> > If you were really organized, you'd maintain a copy of all updated
> > critical files in some directory that isn't part of the "system"...like
> > the sysadmin's home directory tree, or a directory off of /home or
> > something, and always change there then copy into /etc, /var or wherever
> > the files need to live, but all it would take is a small lapse of
> > attention and you'd be missing something potentially important.  If you
> > had such a thing, you could pretty easily copy again after making sure
> > the file contents were still relevant.
> > 
> > How do the experienced folks deal with this problem?
> > 
> > -- Mike B.
> Most configuration files are in the /etc directory.  If you install something
> that stores its config files somewhere else, make a symlink into /etc for it
> so you can just tar the /etc directory occasionally. If you put your /home
> directory on a separate filesystem, it can be left untouched during the new OS
> install.
> When I migrate a server, and I don't have new hardware to install to, I try to
> copy everything to a large enough hard drive or two. Then you can install the
> new OS and take your time migrating.
> If it is a critical server, and you can not afford the down time, You should
> really get another piece of hardware. You can get an older server and move to
> it so the newer hardware is free for upgrade.
> 
> I have done some in place upgrades, but I will usually make a list of all the
> installed rpms with rpm -qa |sort >rpmlist.txt before and after the upgrade,
> and compare the list with a log from a fresh install in a vmware vm. That way
> I can look for orphaned rpms.
> 
> 

-- 
Robert Heller             -- 978-544-6933
Deepwoods Software        -- Linux Installation and Administration
http://www.deepsoft.com/  -- Web Hosting, with CGI and Database
heller at deepsoft.com       -- Contract Programming: C/C++, Tcl/Tk