[SWLUG] Failed Raid 1

Matt Willsher matt at monki.org.uk
Tue Jan 22 12:26:51 UTC 2008


On 22 Jan 2008, at 11:05, Alan Pope wrote:

> On Tue, Jan 22, 2008 at 10:48:32AM +0000, Hugh Saunders wrote:
>> My Parents have windows[1] installed on a RAID 1 (mirror)  volume.
>> The raid controller is built into the motherboard and presents the  
>> two
>> disks as a single volume to the operating system.
>>
>
> Likely not "proper" hardware RAID but "fakeraid" software RAID mostly
> implemented as a Windows driver.

If it's an Intel RAID chip (ICHx-R) then it's software RAID. Even with  
RAID 'enabled' in the BIOS, with the drivers Linux still sees the  
under laying disks not the RAID volume.

>
>> 2. Attach new drives to raid controller in same configuration as  
>> before failure.
>> 2.1 Boot knoppix, try and mount raid volume
>
> Make sure it has dmraid to be able to be able to mount the fakeraid  
> volume.

It's possible you may be able to see the mirror disks as individual  
disks without the need for the fakeraid volume.

>
>> 3. Any other tools, techniques or tips that might help?
>>
>
> In the future don't use fakeraid? Maybe simplify things by having  
> everything
> on one disk and just backup to the other one regularly.


You may find that the disks are ok. I've seen fakeraid fail disks that  
were actually ok due to the bus being saturated.

I agree with Alan - avoid fakeraid, unless backups are  also kept. It  
causes more heartache and problems that it's worth. With a backup you  
don't get the uptime benefits, but you can keep historic data and  
revisions so if bad data is written to your main volume the backup is  
ok, unlike RAID1 systems. If you decide to take this approach it's  
worth getting a bigger disk that the primary OS disk so you can keep  
more data revisions.  If you got a third disk you could use windows  
software mirroring on this I guess thought I'm not sure how reliable  
that it.

Using something like Ghost would give quick recovery.
Often you can't use SMART diags over RAID volumes. Some sort of SMART  
monitoring should always be used where you're worried about disks  
failing.

Another thing to bear in mind if the disks haven't already been bought  
is don't buy the same type of disks from the same place at the same  
time. When they fail, it increases the chances of them going around  
the same time.

The order of RAID reliability goes: fake RAID < software RAID <  
hardware RAID.


My person approach would along the lines of the following, assuming I  
didn't have a second PC:

- Leave original disks as is. Connect one of the new disks to a free  
port on the MB.
- Boot Knoppix.
- Check SMART status of disks with smartctl.
- See what dmraid makes of it. If dmraid sees it, see if it can be  
mounted (read only)
- If it can't be mounted, remove dmraid. Can the underlaying devices  
be seen? If so, try and mount them (ready only)
- If they can be mounted, copy the data off to the new disk (this can  
be the new backup disk) and to the USB drive just to be sure.
- Shutdown and remove all the disks. Connect the disk for Windows.
- Install Windows and setup drivers etc.
- Shutdown and connect a backup disk (or just plug in the USB disk)
- Copy data back to Windows disk
- Install Ghost and create a snapshot to the backup disk.
- Set up regular backups.

If the original disks were ok I'd consider using fakeraid1 on two of  
the disks and using the others for backups. Never use fakeRAID5! ;)

Matt










More information about the Swlug mailing list