SQL and Raid rebuild question

[brendy]

Having seen this problem a few times now, does anyone know how the process of an array rebuild occurs?
Hardware is: Hp DC7700 minitower , adaptec 1220sa 2 port sata card with 2 x 250 WD drives, windows xp sp2, sql msde.
The array shows up as degraded, a few simple tests show the drive as working. A rebuild is initiated from the adaptec software and on a reboot the sql database has corruptions. Am I right in thinking that the sql database is copied to the mirror drive live rather than locked so when the pc is rebooted the differences in data causes the database to **** itself? As a test, I made a backup of the database just as the rebuild was starting off thinking that I would be making a backup of the working drives' DB files and the second drive wouldnt be back online untul after the reboot.
In a nutshell, what are the stages of a mirror rebuild and what changes when the machine is rebuilt?

 

[rpstewart]

The system shouldn't be able to access the two drives independently - they're presented to the system as a single device.

I'm mulling scenarios of uncommitted transactions being rolled back on one drive etc but it still comes back to the point that the good drive should be OK at all times and that should be the source of the rebuild. Then there's the question of how the rebuild deals with changing data as it happens, that will be dependent on how the driver implements the rebuild functions so it's difficult to comment on.

 

[brendy]

I have decided to try a new raid card, something with a bit of cache and leaning towards the mid table cards rather than these cheap silicon image rebadged cards. Cheers. We have found out that if we attempt a rebuild while the sql engine is still running it corrupts the database, if its turned off before the rebuild its ok. Going to have to see how it goes.

 

[brendy]

Raptors are in and running well, I also noticed disabling the index service and disabling the page file, the machine started running smoother.