On Monday June 30, bo@anthologysolutions.com wrote:
> Hi Neil & Raiders,
>
>
>
> I am not quite sure if we really need to resync the array at the time of
>
> RAID 1 or RAID 5 array creation. I tried and tested without RESYNC by
>
> modifying "case1" and "case2" from MDADM source, I could not notice
>
> any problems or issues.
>
>
>
> So I want to know from you what is the reason for these codes, and
>
> what you experieced from without RESYNC.
>
>
Others have answered with very relevant arguments for an initial sync.
For raid1 it is not absolutely essential, but for the reasons others
have mentioned it can be a good idea.
For raid5 it is essential: you do not have any guarantee of data
integrity after a drive fails without it.
Consider:
You start with a raid5 array where all the parity blocks are wrong.
You write to a block on disk A and update the parity using a
read-modify-write cycle. i.e.
pre-read A. pre-read Parity.
subtract old A from parity. Add new A into parity
write new A, write new parity.
The parity block is still wrong with respect to all other data
blocks in that strip.
Now drive A fails. You have lost the block that you wrote there.
I would not be against adding a flag to mdadm which caused it to
create a raid1 array without any initial resync. I would be against
allowing it for raid5.
NeilBrown
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
