Re: preparing for sun server crash

From: Scott (woulduno_at_gmail.com)
Date: 11/02/05 

Date: Wed, 02 Nov 2005 14:24:20 -0600

I personally use a combo of snapshot and ufsdump. Snapshot gives me my
idle file system and ufsdump of course creates an image of the file system.

I then store them on a network share some place. In the event of a major
failure where a full restore is needed I boot into single user via a CD,
configure the network, and mount the NFS share.

I keep a copy of my VTOC and vfstab on my backup server so I can easily
cat the VTOC to fmthard, assuming you replace with a matching disk.

I will then do a ufsrestore of my file systems. I keep a copy of my vfstab
that has been modified to remove any metadevs. I have yet to figure out
how to backup my metadb's and have a successful restore.

Once the restore is complete I do an install boot, reboot, and rebuild any
metadevs while the system is running. One more reboot will be needed after
you do you initial meta setup.

Here is the script I use, found orig on the net some place, to do weekly
full backups over ssh to a remote server.

#!/bin/ksh
LOGFILE=/tmp/snap.log # Log
MAILDEST=someone@yourdomain.com # Who to send ERRORS to
ID=backupid
SERVER=archive
HOST=`/bin/hostname`

####### Root (/)
BS=/bs # Location of backing store

SnapRoot=$(fssnap -F ufs -o backing-store=$BS /)
if (( $? != 0 ))
        then
                print "Failure to create snapshot: /" >> $LOGFILE
                cat $LOGFILE | mailx -s "Snap Alarm" $MAILDEST
        exit 1
fi

ufsdump 0f - $SnapRoot | ssh -l $ID -C $SERVER "gzip -9 - > \
/full/$HOST-root.ufs.gz"

cd /
fssnap -d /
rm -f /bs/snapshot*

Your backing store can not reside on the same partition you are trying to
back up...

I then have rsync setup to grab changed files during the week.. The backup
server has it's own internal rotation to maintain 7 days worth of
incremental backups.

Over all I find this to be a pretty fast restore process, limited only by
the speed of the network. I have successfully done restores over a WAN,
although a little slower than local network.

You could do this same process using a tape drive, but I find them to be
too slow. I prefer this method, and I then have the backup server dumped
to tape via Veritas to maintain our required retention.

Our backup server is a low cost Linux machine with cheap SATA storage and
a tape library attached to it. And by only needing one Veritas license we
save a bit of cash also.

Scott

Relevant Pages

  • Re: VERY VERY URGENT !!!! legato networker old index
    ... Your backup plan MUST ensure that everything you may need is backed up and that the tapes will be retained at least until replaced with a fresh backup! ... I would not rely on a product that does not write to tape EVERYTHING needed to restore a file system. ... When I ship a backup tape off site, I expect to have sent everything necessary to restore that file system even if I have to do the restore in a disaster recovery hot site. ...
    (comp.unix.solaris)
  • Re: HELP PLEASE... (recovery from corruption)
    ... Are your event logs full of "The file system is corrupt" or other ... It sounds to me like a file system corruption. ... metal restore of last night's backup to reformatted disk partition. ...
    (microsoft.public.windows.server.sbs)
  • Re: Backup and Restore of Linux file system
    ... > Thickie Windows user question. ... Windows x file system, tape drive, ntbackup, ... > backup drive c and system state/reg. ... restore drive c and sys state. ...
    (comp.os.linux.misc)
  • Re: Backup tools too slow for LTO-3
    ... have to restore your backup server before you can restore anything else! ... you have to have a a backup server in a DR situation. ... involved in a full client restoration. ...
    (comp.unix.solaris)
  • Re: Backing up an FS
    ... > I have gone through Google groups and searched for good information on ... > how to sucessfully backup and restore a file system. ...
    (comp.os.linux.misc)