Data Storage - General Question to RAID 5

Author

General Question to RAID 5

Krausse Martin

2005-10-07, 7:46 am

Hello, I'm a new user of this newsgroup.
Who can tell me, how a RAID 5 knows, that a hard disk drive is
or goes failure? Because, to recover Data with parity checksum, you
must know, what number of hdd is failure!
I don't find any useful information in the www.
Does a RAID System checks the SMART Informations of a HDD
or does he reads out ERROR Register?

Thanks
Martin Krausse

brufnus

2005-11-28, 10:55 pm

RAID 5 doesn't know anything about the drive's internal status, rather it won't be able to read a failed drive, of course. The checksum is not written to the same drive every time (this is done in RAID 4), but spread across all members of the array.

Thus, if a read error occurs on a member of the array, it simply calculates the missing data by means of the remaining data read from the healthy disks.

If the data's checksum is on the dead drive, it won't be needed - the rest of the drives then actually just looks like a RAID 0 array.

On the other hand, if the checksum for the given data is on a healthy disk, it can use this checksum to calculate the missing data which used to reside on the failed drive.

So it's not actually a matter of the RAID 5 engine being able to tell about the drives themselves, rather it's inability to read from a failed drive, of course.

Example: you write a file to a 4 drive array (without spare), which contains the number "12".

The first disk contains a value of 2, the second contains a value of 4, the third 6, and the last drive receives the checksum which in my example is the sum of the others, namely 12.

If the last disk with the "12" dies, nothing happens except that there's no redundancy anymore. The array can still read the actual data.
On the other hand - if the drive with the value "4" dies, the RAID engine starts working and subtracts the other 2 values - 2 and 6 - from the last drive's value 12, thus it restore the missing value 4.

This is a very simple description, but describes the principle of how it works. So as you see, the RAID itself don't care which drive is unreadable, it simply receives a read error from the kernel driver as any other filesystem would.