I don’t know if there’s some kind of stages to learning a thing the way there are to coping with a thing, but one of them might be when you can help out somebody who’s way ahead of you in your field, and who’s helped you out quite a bit in the past, and they’re glad for the result. I should add that it feels pretty good, too.

Update: For those of you playing along at home, the crucial juju was found in this command:

dd if=/dev/hdc of=/dev/hdd bs=1k conv=sync,noerror


  1. Posted October 28, 2004 at 10:41 am | Permalink


    I’ve never really considered myself a ways ahead of you in your field, I’ve always just thought I was overspecialized in a couple of areas convenient to you.

    Lord knows I’ve needed a lot of help in my day, something to which some of your other readers can certainly attest.

    Welcome to the community, pass some karma, there’s plenty to go around.

  2. Mike Hoye
    Posted October 28, 2004 at 6:28 pm | Permalink

    I’d like to hear some more about that second or third steps, the extracting the copied superblock and the partition table fudging.

  3. Mike Hoye
    Posted October 28, 2004 at 6:31 pm | Permalink

    I should that add that I’m OK on the kitten front for now.

  4. Posted October 29, 2004 at 10:02 am | Permalink

    Well, the “supported” method according to the ext2/3 documentation is to use to guess the partition table. gpart was about as useful the swearing that I tried first. (scroll to fixdisktable) was also reasonably useless.

    I have confidence that would have been somewhat successful, but I wasn’t able to try it. Apparently, the author doesn’t want people using it (and says so) becuase gpart is “better” — so nobody has a precompiled statically-linked executable available. I’d compile it, but, I have no compilers on any of my systems – except the one that died. (security measure)

    I was successful, though, in using to find the superblocks and get the information I needed for the partition table. It returned the start/end of the partitions as I would need to feed in to fdisk – so long as I subtracted one from each number it gave back, because it numbers from one and fdisk numbers from zero.

    I built the required partition table (with a few failed attempts thanks to the 1-vs-0 thing) until the new disk gave the exact same numbers in testdisk as I got from the old disk. I then archived the partition table from the new disk with sfdisk -d /dev/hdb > /tmp/sfdisk-d.hdb

    dd if=/dev/hdc of=/dev/hdb bs=1k conv=sync,noerror

    Then I restored the partition table with sfdisk < /tmp/sfdisk-d.hdb

    Then I tried booting. Which didn’t work, of course, because my 1GB dumpster-provided temporary boot disk had RH9 on it, as did the new 120G disk.

    Whenever you put two disks on which RH9 is installed in to the same system, they don’t boot. This is because the rocket scientists down there use “LABEL=/” to identify the root partition in /etc/fstab instead of something sane like “/dev/hda1” — so a quick boot floppy later, a kick in the pants to the fstab, and I was back in business.

    Now, once it’s all spooled to CD, I’m going to zero out the bad drive and take it back to the store I got the replacement from for a refund because they’re the exact same model and they have no record of the serial number of the new drive they sold me. I will tell them it’s bad and get a refund through their 30-day policy, and they’ll RMA it to the manufacturer instead of me doing it myself through the 1yr replacement-only warranty of the bad drive.

    And I’ve added Maxtor to the list of manufacturers I’ll never buy again. Just like I did with the wasted Fujitsu that this Maxtor replaced, and the wasted Hitachi that it replaced, and the wasted Toshiba that it replaced. I think that leaves just Western Digital, Seagate and IBM that I’ll buy from.

    I’ll take the money I get from the refund, and buy a DVD burner so I don’t have to back up 120G by CD again. No tape, thanks, I’ve had my fill of investing in magnetic media.

  5. DQ
    Posted October 29, 2004 at 2:22 pm | Permalink

    If you want to get an idea of how reliable a drive is, check out

    As for CD backup’s, when the cost per Gig ( less than a buck a Gig : $160/200G = $0.8 / Gig) became cheaper than a reliable brand CD (one that I would trust to do backups), I just bought a drive and used it to backup things. DVD’s are another story since the price per gig is cheap for high quality DVD’s. See for which DVD’s to buy. For Maxells, you will be looking at $1.00/DVD which is < $0.25 /Gig.

  6. Posted October 29, 2004 at 3:18 pm | Permalink

    Which is? Which is? The suspense!

    I do indeed use a hard drive for my primary backup — that was the one I lost. All the data, save for 80G of MP3s or so, was still intact and on the primary servers, I just lost all the backups.

    As for the backups, I also don’t want to trust a single media type. We’re talking about 3800 digital pictures I’ve taken, important mail archives, my websites, etc. No way in hell do I want to lose those, so I want to do a daily backup to this backup server, and spool the new MP3s to CD every month and the new pics to CD every month, etc.

  7. Mike Hoye
    Posted October 29, 2004 at 3:23 pm | Permalink

    If you’re willing to risk a coaster or two, it gets even cheaper than that.

    As in, about seven cents per gigabyte. Sweet Jeebus.

    (DQ said “<“, not the more correct “&lt;”; fixed.)