Link to home
Start Free TrialLog in
Avatar of CompProbSolv
CompProbSolvFlag for United States of America

asked on

SBS 2008 RAID recovery

I am cleaning up after a SBS 2008 server that has had a RAID problem.  It has two drives that are mirrored from within Windows (this will get changed).  As best as I can tell, the system was shut down improperly (power lost, battery backup expired).  After the shutdown, it would not boot from the RAID array (FTDISK_INTERNAL_ERROR message).  The system would boot if the secondary plex was selected.

The problem is that this (not sure what else may have been done) resulted in two drive letters, G: and C:, where G: is drive 0 of the (no longer) mirrored drives and C: is drive 1.  G: is marked as System, Boot, Page File, and Crash Dump.  If I remove the second drive and try to boot from G:, it starts to boot, but gives a message that Directory Services will not start because of a "file not found".  If I remove the first drive and try to boot from C:, it gets nowhere.  I believe that I got a "boot device not found" message.

I've browsed the folders and it appears that the current data is on C:.  I'm not certain about the Exchange or SQL info (yet), but the basic documents on C: are newer than the ones on G:.

My goal is to replace the drives with two new ones, use the controller RAID to make a single mirrored drive array, and restore the system to it.  My problem is that the system doesn't work with just C:.  How should I resolve this?

I tried booting the SBS 2008 installation DVD to do a repair.  I load the drivers and it doesn't find any system.  If I go to the command prompt and run diskpart, the two drives are listed as "Foreign".  The other drives on the same controller are visible, so I'm concluding that these are the correct disk drivers.

I should note that the system does boot with both drives in place if one chooses the Secondary Plex.  Windows seems to be working just fine.
Avatar of R. Andrew Koffron
R. Andrew Koffron
Flag of United States of America image

check with the manufacturer make doublt and tripple sure you've got a a solid backup copy on at least two sets of media that aren't either of the drives in question.

than typically in this type of situation, you just completely remove the failed drive, break the mirror set, than put (I always do a new drive since the bad one failed once already, but alot of people seems to think RAID sets break on their own and it's ok to re-use the failed disk, you can probably tell what I think of that logic) a new drive in and setup a new Mirror.

if the disk won't break in disk manager take a look at this article and http://support.microsoft.com/kb/969749 and consider calling Microsoft, I've found the professional services to be reasonable about refunding cases for things that are their fault.
Avatar of CompProbSolv

ASKER

Keep in mind that the mirroring is done within Windows, not in the controller (a bad situation).

As I mention in the original post, if I remove either drive, the system will not boot.  This prevents me from installing the replacement drives.

You suggest "check with the manufacturer".  For what in particular?

Isn't the fact that the drives show up as two separate drives (C: and G:) clear evidence that the mirror is already broken?
I'd agree that an additional drive letter is a concern, could this have come about another way? A configuration change maybe??
If windows is working well when booting from the second plex, why not use that partition to migrate to your preferred setup..
I'm surprised you are running a SBS solution from a single partition, and especially on s/w raid. It sounds like it is decent hardware though and with a little work it could be kosher..
I expect you must break the RAID 1 first via boot floppy (hope you still have floppy disks)

See http://www.bauer-power.net/2009/07/fun-with-software-raid-on-windows.html
Can you explain how you "browse the folders" and "windows is working fine" if the system will not boot?  Your post is a bit confusing in this regard.

Additionally, the easiest way out of this is to install two new drives, use the hardware raid controller to mirror them, install Windows from the cd/dvd, add the current Windows service pack and restore from backup.  This is probably SATA, and drives are very inexpensive compared to down time.
@debuggerau:
re: "additional drive letter"  I cannot be certain about how we got here as I inherited the problem.  What I described as the history is a combination of was was conveyed to me and what the situation looks like now.  I do know that the system was set up with software RAID initially.
re: "why not use that partition..."  That is exactly what I would like to do.  My problem is that I cannot get the second plex to run on its own.  The first drive is listed as "System" and I think that is what is lacking on the second drive.
re: "SBS solution from a single partition"   Would the present solution be any different if the installer had created a separate OS and data partition?  The data (static files) are not at risk here.  The difficulty with starting over has to do with SQL, Exchange, Active Directory, etc.  Though I could make backups of the first two, reinstalling and (more importantly) reconfiguring Windows and applications is not trivial.

@mpfister: the RAID is already broken as evidenced by the separate drive letters, is it not?

@ fl_flyfishing: In my original post I stated: "the system does boot with both drives in place if one chooses the Secondary Plex".  That is how I can "browse the folder" and "windows is working fine".  The "will not boot" refers to instances where I try to remove one or the other drive.

"easiest way".... that could get rather complicated.  The recent backups are all after the dual drive letters and I see no way in Windows Server Backup to handle a system restore to just one drive.  As stated, my goal is to replace the two drives with a new pair, properly mirrored.

Any thought on if/how the following could work?
Using Windows Server Backup, back up the system (C: and G:) to a temporary drives.
Set up the new drives as a properly mirrored pair.
Restore C: to the new pair, G: to another temporary drive.
Remove G:  (I'm presuming C: will not boot, just as it will not now when by itself)
Do a repair install (?) of SBS 2008, preserving the current settings and installations.
ASKER CERTIFIED SOLUTION
Avatar of R. Andrew Koffron
R. Andrew Koffron
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
You need to tell Windows that you do no longer want a RAID 1, therefore you must break it.
@Harel66:
Thanks for the reminder about backing up.  Yes... I have done it, but the useful backups are only after the drive letter splitting.
re: "boot to your 2008 CD", note my original post:
tried booting the SBS 2008 installation DVD to do a repair.  I load the drivers and it doesn't find any system.  If I go to the command prompt and run diskpart, the two drives are listed as "Foreign".  The other drives on the same controller are visible, so I'm concluding that these are the correct disk drivers

How do I make it boot correctly under these circumstances?  The recovery environment seems to be no help here.
@mpfister:  3 pieces of evidence imply to me that the mirror IS broken:
1)  Windows shows the two drives as two independent drives with different drive letters
2)  The Disk Management GUI does not show the drives as mirrored
3)  Diskpart does not show them as mirrored

Doesn't this imply clearly that the mirror is already broken?
1st off if the drives are split, the one that isn't "c" is a redundant copy, that's serving no purpose(except a backup, and confusing the issue).  if it was a part of a Mirror.

Do your self a favor and stop worrying about the second(failed) drive pull it out and consider is a backup/fall back.

Get a New Disk.

If the Machine has a hardware RAID or SCSI controller your probably going to need a Driver to load while the Installation disk loads I'd call the hardware manufacturer and get the best driver and instructions.

@Harel66:
Just to be clear, the first drive is the failed one, the second one is the good one (C:).   That aside, here's my dilemma with your suggestion.  I"m not "worried" about the failed drive.  I am trying to replace both of them.  The question asked is how to accomplish that.

If I remove the first drive (G:), as I stated above it won't boot.  Adding a second (non-formatted, empty) drive shouldn't change that.  If I then boot from a SBS 2008 DVD and install the proper disk drivers, as mentioned above, it doesn't see the system on the good drive and therefore won't let me repair it.

I have confirmed that I have the correct driver and instructions by attempting a new OS install on a new disk.
Can you show that it is a mirror, as it seems doubtful
@debuggerau: I'm not sure what you are asking.  At this point, I have no indication that Windows thinks it is a mirror.  The two drives show up separately in Disk Management and now have slightly different data (C: has the newer data files).
I don't even know what to say at this point,

a mirror is a Mirror, it doesn't really matter if it's a software mirror or a hardware mirror,

if the machine is seeing 2 different drives with nearly identical data, the older more outdated drive should be removed, cause it's not a mirror anymore.

it's a broken failed redundant drive, that is getting more worthless every minute it's not in a set. it won't link back up unless you re-initialize a mirror using it as a the second drive in a set.  now a week later it's essentially a paper weight,  you've got multiple backups, remove the drive from the machine. triple and quadruple check that you've got all the data you can possible have back upped up,  and replace the older one,  (keep it for sentimental reasons, for an extra backup "just in case", or format it ctrl+click in drive manager, and add it as a new mirror) but really move on, get it out of the machine or force it back into the mirror. all your doing is confusing the situation.  as soon as you made the failed mirror boot back up, and had a tape backup, the 2nd half of the mirror became worthless.  

Personally I'd assume it's a bad drive and replace it.  I'm not a big fan of software mirroring but it's better than no RAID at all.
just realized I deleted a paragraph accidentally.

call the manufacturer to get the right drivers. make sure to try the repair without the bad drive in the machine.  whatever sequence loads the OS on new disks should be able to repair the remaining drive, (or you've still got drive problems) OR it's possible both drives are damaged/corrupted.  so also try a full restore onto the new system.  even if you need to do a full restore this problem has been drawing on too long.  you always have the copy to fall back on, just do what you need to do, this limbo stuff is no good for you or your clients.
@Harel66:

re: "a mirror is a Mirror" - There really is a difference here between hardware and software mirroring.  With hardware (controller-based) mirroring, all data is on both drives.  When using software (Windows-based) mirroring, the drives are not quite identical.

Regardless, we are on the same page as what should be done.  As stated, I AM trying to replace both drives (and change to controller-based mirroring).
The issue with the full restore is that I'm using Windows Server Backup and it won't allow you to restore to a drive that is smaller.  The difference in size here is pretty minimal as the new drives are the same size as the old ones, but by moving to controller-based RAID, there is a small amount used for the RAID metadata.

I'm trying to use different software now and may be making progress.
Finally got it resolved!

I got to the solution with help from what Harel66 mentioned in 35987284.

There were two key steps that I needed to perform, prompted by the comment "use the recovery environment to make the good disk boot correctly":

1) in DiskPart, use the Import Disk command to change the disk from Foreign
2) in DiskPart, adjust drive letters, then use bootrec to create a new BCD.

I used Active@ Disk Image to make a backup to a single internal HD, then created a RAID 1 controller-based array with two new drives, then restored the image from the backup to the new array.
More details about how to make the disk boot (specifically, the "import disk" and using bootrec to rebuild BCD) would have been a big help.