On Tue, Dec 21, 2010 at 6:47 AM, Eric Martin <eric.joshua.martin@gmail.com> wrote:

On Sun, Dec 19, 2010 at 11:15 AM, Theo Van Dinter <felicity@kluge.net> wrote:

Did the second drive completely fail or is it just a bad block? Does
the array work at all? If the disk is marked bad (RAID metadata),
then maybe you can mark it as ok, get the array online, and copy off
as much data as possible (starting with the most important files
first, of course).

Bad blocks and I think it's still marked ok (see other email thread with mdadm data.

You may be able to dd the data per disk, skipping any bad blocks, to
another disk and then possibly be able to access some of the array
that way.
Done

And of course ... you do have a backup of that data somewhere, right? :)

Yep, just not as current as I want it (within a few days).

In general, some rules of thumb are:
- RAID is not a replacement for backups
- RAID 5 becomes more problematic as disk sizes + number of disks
increase, look at something like RAID 6 (raid-dp, raidz2, whatever
your implementation calls it) instead.
- Be aggressive about fixing bad disks in order to shorten the
catastrophe window.

The purpose of raid in this machine was to make it faster and fail gracefully (of course it not mentioning that a drive or two failed isn't living up to that promise). I have rsnapshot running every 4 hours, but the last few days of the server running it wasn't getting everything and I didn't realize it. The goal right now is to understand software RAID better while trying to get my data back in the window of no backups. If I can't get it running tonight, I'm throwing in the towel and restoring from backup.

On Sat, Dec 18, 2010 at 1:03 PM, Eric Martin
<eric.joshua.martin@gmail.com> wrote:
> It's raid 5, and I can't recover the group because of a hd error on two
> disks. I wanted to see if I could move that data somewhere else to force a
> rebuild but it looks like a no. Oh well, to the backup drive.
>
> On Sat, Dec 18, 2010 at 1:00 PM, Jason Couture <plaguethenet@gmail.com>
> wrote:
>>
>> What kind of raid? If it was raid 5 I don't believe you can rebuild the
>> array unless you have 2 surviving disks, so unless you can somehow recover
>> one of the other two it would seem you are sol. Been a long time since I've
>> done anything but 0, 1 and 0+1 so I could be wrong.
>>
>> Sent from my android wireless device.
>>
>> On Dec 18, 2010 12:48 PM, "Doug Mildram" <dmildram@gmail.com> wrote:
>>
>> You're hosed (i think) unless
>>
>> -- you're rich and you send the disks to a shop for megabucks (they may
>> start with a trivial purchase of 2-3 identical model disks)
>>
>> -- OR IF your problem was NOT the actual platters/heads, YOU obtain 1-2
>> same-model disks, swap the board from a replacement disk onto your hosed
>> disk ( w/ sm torx tool and brave fingers )
>>
>> I've not done this myself yet.
>>
>> I HAVE had "stiction" problems where (after a power out or a "rest" so a
>> disk bearing/lubricant gets a chance to harden) the disk just isnt spinning
>> up. Of course, tapping/whacking it at the right/initial moment will usually
>> cure that and you can feel if it's spinning or not....if you can hold it in
>> your hand ...OR somehow get hands-on during powerup.
>>
>>
>>
>> _______________________________________________
>> Wlug mailing list
>> Wlug@mail.wlug.org
>> http://mail.wlug.org/mailman/listinfo/wlug
>>
>>
>> _______________________________________________
>> Wlug mailing list
>> Wlug@mail.wlug.org
>> http://mail.wlug.org/mailman/listinfo/wlug
>>
>
>
>
> --
> Eric Martin
>
> _______________________________________________
> Wlug mailing list
> Wlug@mail.wlug.org
> http://mail.wlug.org/mailman/listinfo/wlug
>
>

_______________________________________________
Wlug mailing list
Wlug@mail.wlug.org
http://mail.wlug.org/mailman/listinfo/wlug

--
Eric Martin

_______________________________________________
Wlug mailing list
Wlug@mail.wlug.org
http://mail.wlug.org/mailman/listinfo/wlug