SUMMARY: Backup 500 workstation to DAT drive on Alpha 2100A

From: <patrick.hayes_at_ast.com>
Date: Mon, 20 Oct 1997 15:45:57 -0800

     Original message :
     
     
     I don't have a DAT drive on the Personal Workstation 500. Is it
     possible to do a vdump backup of the workstation over the network to
     a tape device on a Alphaserver 2100A (DLT or DAT) ?
     
     If so whats involved ? on the 2100A the DAT drive is /dev/rmt0h, can
     I try to nfs mount it and then refer to the device as
     /dev/rmt0h_at_2100_hostname ?? I don't think this would work ! >
     Also, as an aside, can I do the same for a SUN Solaris workstation
     ?
     ----------------------------------------------------------------------
     Thanks for all the prompt replys. There are several ways to skin a
     cat, but the solution I have gone with is vdump/vrestore piped into
     rsh.
     First, I modified the .rhosts file on the server to allow root on the
     workstation to connect without a password.
     Then, from the workstation
     
     vdump -0 -F 64 -b 60 -f - / | rsh juliet dd ibs=60k obs=60k
     of=/dev/rmt0h
     
     Simple really ! This worked fine,it was fairly slow. We are looking at
     ways of speeding this up across the network by changing the IP address
     of the workstation so that its on the same backbone as the
     AlphaServer.
     Now all I need to do is test that the restore works ! .
       
     Other solutions/ideas mentioned include gtar and rdump/rrestore for
     ufs,
     
     Thanks to
     T o m L e i t n e r
     hl_at_tekla.fi
     Jason.
     Mark R Huehls
     Alan
     Lucia Gonzalez
     Robert L. McMillin
     
     Regards
     Patrick
     
     Below are the replies.
     ----------------------------------------------------------------------
     
     
     Sure. The trick is to pipe the vdump output over the network to the
     other workstation. I use something like this:
     
     # vdump -v -0 -F 64 -b 60 -f - /home |
     rsh remote_host dd ibs=60k obs=60k of=/dev/rmt0h
     
     provided that you have "rsh" access to the other workstation as root.
     
     Another way would be to use "gtar" with
     
     gtar cvf remote_host:/dev/rmt0h /home
     
     Hope this helps -- Tom
     
     T o m L e i t n e r
     ----------------------------------------------------------------------
     
     vdump 0f - /"dir" | rsh a2100 "dd of=/dev/rmt0h"
     
     Of course you need to set up .rhosts first.
> Also, as an aside, can I do the same for a SUN Solaris
     workstation ? Sure. But the tapes have to be read back the same way to
     the same machineplatform.
     rsh a2100 "dd if=/dev/rmt0h" | ufsrestore xf -
     
     But if you want a goo backup solution, have a look at DECnsr (i.e.
     Legato Networker
     for solaris). Your 2100 should a server license ready, all you need is
     a client license
     for you networked clients.
     --harald
     
     hl_at_tekla.fi
     ----------------------------------------------------------------------
     If you have ufs filesystems you can do use rdump/rrestore -- look in
     the man pages. Solaris supports the rdump command so I would use that
     for the sun machine to the alpha box -- you may need to use some
     options to support the different plateform but it should work.
     
     If you are using AdvFS then you will have to use vdump, as your mail
     states. You could nfs mount your filesystems on the remote box and
     use vdump -- as vdump is filesystem independant you wont have any
     trouble backing up a nfs. If you could arrange setting up some
     /.rhosts (making your root a trusted user on the remote system) files
     you could use a rsh command like this to do your backup (and the other
     way round for the restore). Test it to your satisfaction before you
     commit to the backup stratergy
     
     vdump 0uf - | rsh remote_machine dd of=tape_device_file bs=32k
     
     I think 32k is the output of the dump command.
     
     
     If you have ufs filesystems you can do use rdump/rrestore -- look in
     the man pages. Solaris supports the rdump command so I would use that
     for the sun machine to the alpha box -- you may need to use some
     options to support the different plateform but it should work.
     
     If you are using AdvFS then you will have to use vdump, as your mail
     states. You could nfs mount your filesystems on the remote box and
     use vdump -- as vdump is filesystem independant you wont have any
     trouble backing up a nfs. If you could arrange setting up some
     /.rhosts (making your root a trusted user on the remote system) files
     you could use a rsh command like this to do your backup (and the other
     way round for the restore). Test it to your satisfaction before you
     commit to the backup stratergy
     
     vdump 0uf - | rsh remote_machine dd of=tape_device_file bs=32k
     
     I think 32k is the output of the dump command.
     
     Hope this helps,
     
     Jason.
     
     Actually patrick -- its 64k you want to use as 32k will probably give
     you an error message like so...
     
     vrestore: unable to read from device '/dev/rmt0h'; incomplete read
     32768 < 65536
     
     but 64k works a treat.
     ----------------------------------------------------------------------
     set up network file system and do a remote backup. the 2100 would be
     the remote device. setting up nfs is the real trick.
     
     Huehls Mark R
     ----------------------------------------------------------------------
     The common method for doing vdump to a remote tape is something like:
     
     vdump <options>f - filesystem | rsh remote dd of=tape-drive
     bs=block-size
     
     You'll need the block to get dd(1) to write a reasonable block size to
     tape and reblock whatever comes off the pipe. Then you'll need to
     experiment with what options vrestore will need to get data it likes
     so you can restore the backup. I've read of other using something
     like this and not checking to see that they could restore until it was
     too late.
     
     re: NFS served devices.
     
     You're right, it won't work. NFS serves file systems. When a special
     file is served via NFS, the local host interprets the device number.
     
     
     Alan
     ----------------------------------------------------------------------
     We had the same problem. We have 9 AlphaStations 255 without local
     tape units, and an AlphaServer 8400 with a DLT and a DAT. We used to
     dump everything on a large file on each workstation (with vdump) and
     then just copied the 9 files to a NFS exported directory on the 8400
     (/backup), from which we just did a
     tar -cvf of /backup to the /dev/nrmt0h.
     
     But this was disastrous, when someone wanted to recover a file, we had
     to do a painful job of unpacking everything, blah, blah.
     
     So we came up with a solution, a vdump|dd backup through rsh.
     
     Steps:
     
     1) on the AlphaStation console:
     
     # /sbin/shutdown -h now (single user, we want the backup to be
     perfect)
     
>> boot -fl s
     
     # /sbin/bcheckrc (check and mount all filesystems)
     # /sbin/init.d/inet start (so that the system knows its name) #
     /usr/sbin/inetd /etc/inetd.cccfc.bck
     
     where /etc/inetd.cccfc.bck is a file that just contains 1 network
     service: rsh
     ----------------------------------------------------------------------
     -
     shell stream tcp nowait root /usr/sbin/rshd rshd
     -l
     ----------------------------------------------------------------------
     -
     
     Thus we leave the machine accesible with rsh by root from the server,
     but no other network services open, and no rsh for the rest of users
     (flag -l)
     
     2) On the AlphaServer console:
     
     # mt -f /dev/nrmt0h rewind (or fsf #n if it's not the beginning of the
     tape) # /backup/script.ksh
     
     
     where /backup/script.ksh does the following (for level 0 backups)
     ----------------------------------------------------------------------
     ---------- ------------
     #!/bin/ksh
     echo "Which machine are you backing up?" read workstation
     
     today=`date +%d/%m/%y`
     logfile=/backup/$workstation.$today.log
     
     time rsh $workstation "/sbin/vdump -0uf - -b 1 /" | dd of=/dev/nrmt1h
     bs=1024 >> $logfile 2>&1
     time rsh $workstation "/sbin/vdump -0uf - -b 1 /usr" | dd
     of=/dev/nrmt1h bs=1024 >> $logfile 2>&1
     
     mailx -s "Level 0 backup ($workstation/$today) to DLT tape"
     root_at_mailaccount < $logfile
     rm $logfile
     ----------------------------------------------------------------------
     ---------- -------------
     
     it works! if you want to recover anything from the backup you just
     have to sit on console on the server and:
     
     # mt -f /dev/nrmt0h fsf #N
     # cd /backup/recover
     # "dd if=/dev/nrmt1h bs=1024"|vrestore -if -
     
     you get the interactive prompt from vrestore and that's fairly easy...
     
     But this strategy is incomplete, since we would like to have the
     backups done during the night, causing the least disruption to the
     users.
     
     Ideally, there should be a rc.4 level on the workstation. This level
     would start bcheckrc, inet, inetd, and kill everything else.
     
     Then I would write a crontab on the AlphaServer, which sent at 2:00 AM
     by rsh an
     
     init 4 to the workstation, waited for a while, and then executed
     /backup/script.ksh.
     
     But I have no idea of how i can develop a level without doing anything
     wrong! Any suggestions?
     
     Thanks in advance,
     
     Lucia Gonzalez
     System Administration.
     ----------------------------------------------------------------------
> /dev/rmt0h_at_2100_hostname ?? I don't think this would work !
     
     Nope, you'd say "vdump -f hostname:/dev/rmt0h /". However, that's
     painfully slow unless you have 100BaseTX on all your machines. I
     wouldn't recommend this for the folks at home anyway, since the rmt
     protocol is slow to begin with once you consider overhead.
     
> Also, as an aside, can I do the same for a SUN Solaris
     workstation ?
     
     Yup, provided you have ~root/.rhosts correctly set up.
     
     --
     Robert L. McMillin
Received on Mon Oct 20 1997 - 17:55:07 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT