[mythtv-users] Consequences of drive failure

[Mike Perkins]

On 03/12/12 21:14, Michael Watson wrote:
> On 3/12/2012 7:19 PM, Simon Hobson wrote:
>>
>> As to copying your drive, my technique is this :
>> Install the new drive in the system along side the existing one - you can now
>> let the system continue running as normal.
>> Partition & format as required, mount somewhere on the filesystem. Use rsync
>> to copy the recordings from the old drive to the new one. You can do this in
>> batches if you want, stopping before the backed has to do any Myth related
>> work (eg recording or serving up frontends).
>> When ready to make the switch, stop the backend, run the rsync copy again to
>> get the replacement drive fully up to date, make any changes that might be
>> needed to fstab etc, shutdown, remove the old drive and start up.
>>
> I would add the new drive, create filesystem and mount, add the new drive to the
> default storage group, and remove faulty drive from the storage group, so no new
> recordings will be written to the old drive, (but you will not be able to access
> the recordings on the old drive from within myth unless you create a new storage
> group with the old drive included).
>      Copy the recordings over using rsync, but precede the rsync command with
> "ionice -c3", and let the copy run its course.  Once done, modify fstab to
> include new drive and remove the old drive, shutdown and physically remove old
> drive.
>      Power up, run find_orphans.py to clean up any missing recordings.
>
So, what is going to happen when your rsync operation meets the first defective 
sector?

As it happens I have no intention of attempting any of this with the existing 
server host, (i) I would expose the other drives to potential operator error 
(ii) there are no free SATA slots which means more potential error juggling 
drives around (iii) every time I boot the server mythbackend will start and try 
to do things to the disks.

There is a minor issue I have: the two old drives have the same geometry, the 
two new drives do as well, but the old drives have 512-byte sectors, the new 
ones have 4096-byte sectors. My plan is this:

Remove good old drive, mount with new drive in spare host. Rsync files across. 
Put new drive into MBE slot where good drive was. (Er, fix fstab for new UUIID.)

Remove bad old drive, mount into spare host alongside good old drive (same 
geometry). Use dd_rescue to bit-copy from bad to good. When happy with result 
remove bad old drive.

Put remaining new drive in spare host, rsync from good old drive to new drive. 
Remove new drive and put into MBE.

??? Profit!

-- 

Mike Perkins