RE: Disaster recovery for my Redhat 6.2

["Karl Pearson" <karlp colubs com>]

I've made my living in disaster recovery for the past decade on Unix systems
and have come up with a reasonable, IMHO, plan for recovering after a
significant crash where the disks are trashed.

The principle is that you re-install the OS from scratch mimicking as
closely as possible the setup of filesystems, or making improvements you
couldn't while the system was up. To do this effectively, you need to keep a
copy of a df -k so you can see what was where. If you are very picky that
everything gets back exactly as it was, you might wish to do a tree of your
filesystems. That's the extra mile, however, and my experience has shown
that there's almost always a better way to setup filesystems because after
experiencing your system for a while, you see things you wish you would have
done a little differently.

After the restore of the OS, you then selectively restore etc files,
including conf files, inittab, passwd, rc files, etc. to a filesystem that
will hold them. From there, you verify that nothing in, for example,
inittab, will make the system unbootable if it's included. This is pretty
easy on Linux because the inittab is kept pretty vanilla. That's definitely
not the case on other Unix systems.

I think you get the idea from this, that you don't just do a wholesale
restore of the entire backup. For this reason, I keep a backup of the entire
system on one tape, performed weekly, then do separate filesystems nightly.
I include changing data filesystems in the nightly backup. I also include
/home in case any of my users have made changes to their .profile-like files
(.bash_profile, .bashrc, etc.). I try to make the lives of the users as
secure as possible, but I don't tell them that 8^}. I don't want them to get
cocky.

I also include the major applications in the nightly backup. Not because
they change very much, but because I then know that I can restore config
files if necessary. With many applications, especially RDBMS systems, you
will have to re-install so the links and the pseudo kernel files they need
are compiled appropriately. If you know the applications well enough, you
know what to restore specifically so you don't have to restore everything.
One reason I backup everything in the application areas is because I'm sure
the vendor doesn't tell me everything I need to know to be confident I'm
backing up exactly what I need and no more. I've had occasion to be told by
a vendor support engineer that I need a file that hadn't been backed up. I
learned my lesson on that.

I hope this gives you some idea what the process might be for recovery from
a system crash. Just hope it's not the disks (or the disk controllers which
have a way of corrupting data on good disks as the go down), then all you
have to do is replace the bad part and reboot.

Karl L. Pearson
Senior uniVerse Database Analyst
Senior Unix/NT/Win Analyst
karlp colubs com

-----Original Message-----
From: Anurag Jalan [mailto:a_jalan vsnl com]
Sent: Friday, July 28, 2000 12:37 PM
To: redhat-install-list redhat com
Subject: Disaster recovery for my Redhat 6.2


Hi all,

I take regular backups with taper ( Full .. about 2.0 GB ) to my HP SCSI
dat drive........In this manner..*everything* except the contents of /tmp
is backed up...

I would like to know... how best to restore my present system in case of
a HD crash or something ........

I run SAMBA, WWW ( compiled with MySQL & PHP ), FTP, Sendmail/Fetchmail
and a few other services.........

any pointers would be very welcome............

Oh yes.. i have the boot & rescue disks ready...

Regards

Anurag



--
To unsubscribe: mail redhat-install-list-request redhat com with
"unsubscribe" as the Subject.