[mythtv-users] OT-RAID 5 failure. MDADM experts? Seagate 1.5TB failure
george_mythusers at mari1938.org
Thu Dec 18 13:54:31 UTC 2008
Jake Anderson wrote:
> Format the lot and start over?
> you might not be able to recover it without going over the top.
> If the disks were in use at the time of failure then critical chunks of
> data might have been changed between disks.
My personal experience suggests otherwise. I have 2 RAID5 arrays that
have had 3 separate instances of 1 drive failures and a second drive
with a different value for the 'events' counter between them, and I've
been able to recover every single time.
For me, it would have been much more time to "format the lot and start
over", and restore from my backup, than it would to just take a little
time to do some research on how to recover.
> IE the array is in an inconsistent state and theres no real way to
> reconcile that.
With one drive dead and another drive with a different value for
'events' like Mark has, you can force assemble the array with one
missing drive, fsck the filesystem to make sure it's ok (JFS has never
missed a beat for me) and add in a new drive.
Now, granted - in my situations, I knew that the filesystems that were
on these arrays were not being written to at the time of failure, so
that made success much more likely.
> Only other thought I seem to remember something about needing to pass
> mdadm an option when rebuilding to tell it to make it into a new array
> or some such, might only have been when you moved the array to a
> different host though.
There is no such option, to the best of my knowledge. Have you done
this before? Even explicitly telling mdadm to create a new array, using
the same devices/partitions that were in the array before, mdadm will
basically try to re-create the old array if it sees there was an array
there before. That's what one of the links I posted previously explain.
> When you sort it all out, use storage groups rather than raid, you will
> get much better performance and probably disk life, the seek load is
> going to be drastically reduced.
To each their own. I have nothing against storage groups, and haven't
tried them yet, as RAID5 has been wonderful for me, but that's an
awfully general statement regarding "much better performance". Got any
hard data comparing performance of RAID5 and storage groups? Yeah,
yeah, I know - the infamous RAID5 write performance penalty. Sure, it's
there - but for mythtv, or for storing large amounts of media files, who
cares? Write performance is more than good enough for those use case
More information about the mythtv-users