I see dead drives

Remember the caveat about error checking:

http://blogs.zdnet.com​/storage/?p=162

Some say this is scaremongering (and dated) but it doesn't affect all RAID controllers, it does affect most software controllers (read those that come with mobo's). In other words invest in a decent RAID controller from Adaptec, 3Ware et al and before buying confirm it's URE.

I have seen enterprise RAID arrays fail to rebuild from a URE error and it's not pretty trying to restore data when your phone's on fire from the management!

This is why you need a independent back up system from RAID in case of RAID failure or data corruption.

Faolan wrote in post #6464395
Remember the caveat about error checking:

http://blogs.zdnet.com​/storage/?p=162

Some say this is scaremongering (and dated) but it doesn't affect all RAID controllers, it does affect most software controllers (read those that come with mobo's). In other words invest in a decent RAID controller from Adaptec, 3Ware et al and before buying confirm it's URE.

I have seen enterprise RAID arrays fail to rebuild from a URE error and it's not pretty trying to restore data when your phone's on fire from the management!

This is why you need a independent back up system from RAID in case of RAID failure or data corruption.

This is interesting but it seems a little flawed, remember that in a RAID 5 the data is sliced up between ALL of the drives and then a Parity Bit is calculated by RAID Controller be is software or hardware and placed on the last drive. This in today's RAID systems is not a single drive.

So lets say that in a RAID 5 a drive fails, you replace it and are now rebuilding it. If by chance you hit a sector that can't be read there is still a good chance that the parity bit is available on one of the other drives and not the drive being rebuilt.

This is why I like to allow my RAID Controllers perform a verify on an array often. I do mine once a week. On a 6TB Array it only takes about 4 hours to complete. This makes sure that all of the data and parity is readable and in place. An error will be generated if it isn't. If you start to see a specific drive have errors then it is time to replace it.

At the same time I agree that no data storage method is bullet proof as they all have downfalls and are subject to failures at some point in time. We need to realize these risks (which a lot of people do not) and take steps to minimize them as much as possible to the point that we are comfortable enough and can sleep at night.

Guys, RAID was never intended to replace a backup system.

It's designed primarily to reduce or eliminate the downtime usually required to replace failed hardware, then restore backups. Even if you have a (modest) RAID setup, you still definitely do need backups.

RAID is MUCH more about 'availability' of a system, than it is about data integrity.