[MUD-Dev] fault tolerance and character files

Vincent Archer archer at nevrax.com
Wed May 2 08:58:51 New Zealand Standard Time 2001

According to Kwon Ekstrom:

> I'd say that if you were worried about keeping your files from
> corrupting, use RAID5.  What that does is write 3 bands across 3+
> hard drives, including a parity strip.  Using data from the
> surrounding drives, and the parity strip, it can reconstruct bad
> data, even if a whole hard drive fails (but not 2 drives)

2 simultaneous hard drive, nope. But if 1 drive fails, then a 2nd
drive does, most raid controllers have a configuration mode with "hot
spares" where a 1 drive is kept in reserve, unused, and as soon as the
first drive fails, the spare one is rebuilt from the remaining
disks. Then, you can suffer 2 drive failures without intervention

However, this doesn't answer the original question. All the RAID of
the world will not prevent your data corruption if the power supply of
the raid array decides it's been enough time since the warranty
expired, and that its manufacturer's MTBF record need a little bit
lowering. Then, even an UPS will not prevent the drive shutting down
in the middle of the transaction.

The short answer is, of course: never delete the old data before you
know the new one is on disk. That and a raid array will go a long way
to ensure you don't get "hard" data corruption. Soft (i.e. data
corruption stemming from a software bug) corruption will then come to
dominate as the most probable source of error.

Vincent Archer                                         Email: archer at nevrax.com

Nevrax France.                              Off on the yellow brick road we go!
MUD-Dev mailing list
MUD-Dev at kanga.nu

More information about the MUD-Dev mailing list