Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Imaging Linux Disks -For the clever reader and needy poster

0 views
Skip to first unread message

Rob

unread,
Jul 17, 2002, 12:36:14 PM7/17/02
to
Can this be done?

Scenario: A mail server is running linux. We need to find a solution
to backup/restore the disk in the event of a disk failure. We have a
lot of spare hard drives of various sizes (larger than the source).

What is the favored soluttion in this case? Obviously we'd prefer to
remove the bad disk "pop" in the new one and just boot and go -no
hassle, minimum down-time. Can we do it across the network to a spare
machine (will it do the MBR, partitions and do it periodically)?
Could I "send" the image to a file and somehow re-build it to a new
disk on another machine?

FLAME DISCLAIMER:
Our company does not have a lot of cash, so fancy options are out.
Yes the machine should have been setup properly before it was
installed but it isn't, hence the need for this post.

Thank you for yor time and knowledge!

Rob

Marc Werner

unread,
Jul 17, 2002, 12:49:27 PM7/17/02
to
> Scenario: A mail server is running linux. We need to find a solution
> to backup/restore the disk in the event of a disk failure.

I would recomend setting up a raid (software if you can't afford a hardware
raid adaptor)
and using hot-swap drive bays so you can swap them with no downtime at all.
The raid
should preserve your data and you have no hassle with any downtime and/or
losing mails

>Can we do it across the network to a spare machine

You could probably setup a second machine as fall-back but this would be
very expensive and
I don't know of any software which will do the mirroring aout of the box.

> Could I "send" the image to a file and somehow re-build it to a new
> disk on another machine?

For this I would do a Ghost Drive-Image (eventhough Ghost doesn't support
the compression of
any Linux Partitions at the moment). But this is only if your whole system
blows at once and you don't
want to reinstall all the software (and update it).

Also don't forget to put a big enough backup device (tape?!?) in the server
and perform a backup on a regular basis (if mail is mission critical I would
backup the server every couple of hours)

> Our company does not have a lot of cash, so fancy options are out.

Well it will all cost a bit but when you want to save your data you will
have to pay for it. If you don't have enough money to set up a proper system
and don't have to much treffic go and look for an ISP which will do the job
for you.


Gary

unread,
Jul 17, 2002, 1:26:38 PM7/17/02
to

Marc Werner wrote:
>
> > Scenario: A mail server is running linux. We need to find a solution
> > to backup/restore the disk in the event of a disk failure.
>
> I would recomend setting up a raid (software if you can't afford a hardware
> raid adaptor)
> and using hot-swap drive bays so you can swap them with no downtime at all.
> The raid
> should preserve your data and you have no hassle with any downtime and/or
> losing mails

If you use any type of removable drive drawer you run two
potentially
big risks.

The first is the possibility of spiking the power supply, and
or the
processor mains when you apply power to the drive. This will
force
a BRS if it happens.

Make certain that whatever system you use has hot swap
controllers
built into it to control the power cycle.

The next problem with "drawers" is that they can cause noise
on the IDE bus. Generally if you see timeout or CRC errors
after installing these devices, you found the cause. You may
be able to get by this issue by throttling down the DMA
transfer
rate of the drives by getting the appropriate tool from the
drive
maker. All of the manufacturers have a utility that does this.


Regards,

Gary

nuk

unread,
Jul 17, 2002, 3:19:55 PM7/17/02
to
In article <e6327a65.02071...@posting.google.com>, Rob wrote:
>Can this be done?

>Scenario: A mail server is running linux. We need to find a solution
>to backup/restore the disk in the event of a disk failure. We have a
>lot of spare hard drives of various sizes (larger than the source).

>What is the favored soluttion in this case? Obviously we'd prefer to
>remove the bad disk "pop" in the new one and just boot and go -no
>hassle, minimum down-time. Can we do it across the network to a spare
>machine (will it do the MBR, partitions and do it periodically)?
>Could I "send" the image to a file and somehow re-build it to a new
>disk on another machine?

I would check out mondo (www.mondorescue.org). It was originally
designed for a bootable backup to a CD burner for small systems, but
IIRC, it can also store the back up images on a hard drive, tape,
whatever and use boot floppies. I may be wrong on the specifics, but I
believe the concept is correct.

Your post implied that downtime of a few minutes to say, a half hour
wouldn't be all bad. Perhaps something as simple as installing a spare
drive, and then using something like dd or rsync, depending on the
requirements to mirror the disk. I guess it would kind of depend on
whether you have your data on a separate disk from the system files, and
just want to backup your mail spool, or if its something where you want
to actually image the drive, and then restore it to a different sized
drive. In the former, rsync should work extremely well; in the latter,
dd would work, but if you have an 8gig hd and then image it to a 10gig
hd, you will only have an 8 gig image on that drive, effectively wasting
2 gig (better than no backup at all, though). Again, mondo is supposed
to be *good* at doing something like: backup, nuke the hard drive, repartition
the way you want, and then restore, and adjust your mount points
accordingly. A bit of manual labor and knowledge is required, but
nothing too difficult. If it didn't work, just reboot the mondo backup
in non-interactive mode and it will put you back where you started.
Alternately, instead of repartitioning the old drive, you can swap out
to a new drive as well. The backup/restore from CDR/CDRW is kind of
slow, though, and I have a 16/10/40x drive. The next release of Mondo
is supposed to haved some sort of network (specifically NFS) capability,
so it could be worth checking into.

nuk


--
"It is the American vice, the democratic disease which expresses its
tyranny by reducing everything unique to the level of the herd."

Henry Miller, American author, 1947

Marc Werner

unread,
Jul 17, 2002, 3:31:52 PM7/17/02
to
> The next problem with "drawers" is that they can cause noise
> on the IDE bus.

Using IDE drives in a server is a bit risky anyway. These drives aren't
build for 24/7 work and therefore probably fail more often.
Unless you really haven't enough money (buy a smaller processor) or the data
isn't imporant to you (why should it then be saved anyway) get yourself some
SCSI drives. These are (with the correct drawers) fully hot-swapable.

There's also supposed to be some special raid/hot swap drawers for IDE
drives but I've never tested them.
If anyone got experience with them please post a message.


mjt

unread,
Jul 17, 2002, 4:07:17 PM7/17/02
to
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Rob wrote:

http://www.partimage.org/

- --
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Michael J. Tobler: motorcyclist, surfer, # Black holes result
skydiver, and author: "Inside Linux", # when God divides the
"C++ HowTo", "C++ Unleashed" # universe by zero

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE9Nc14tTveLPAHcDIRAoUvAJ9520B5eGMjt2bpDQ8nVbMhdjOqMwCeP9fU
lVXPR5QAiIpW+UsqRoOeUBg=
=XUZA
-----END PGP SIGNATURE-----

Rob

unread,
Jul 17, 2002, 8:45:42 PM7/17/02
to
Thank you!

Rob

unread,
Jul 17, 2002, 8:45:55 PM7/17/02
to
Thank you!
0 new messages