Re: saving RAIDframe parity accross a panic

[Greg Oster <oster%cs.usask.ca@localhost>]

Edgar =?iso-8859-1?B?RnXf?= writes:
> > There are plans to make it not have to check as much parity 
> > after a panic, but right now it has to verify all parity...
> My point is not as much that the machine spends hours re-writing the parity,
> but what happens if I have a real I/O error. If the machine panics leaving the
> parity dirty (or more precisely: leaving an indication on the components 
> saying
> the parity is dirty where it actually needn't be), then I'll be losing data.

When the first (time-wise) partition of a RAID set is opened, the 
first thing that is done is to mark all the components as having 
"dirty parity".  This is done before any writing of data takes place.
If you then lose a component (e.g. disk dies) then at that point the 
system considers all remaining parity as valid -- this is fine, 
because the parity was valid before, and it's been kept up-to-date.
If the system now panics, the component labels will say that the 
parity is dirty, but since a component is missing it will be forced 
to use whatever parity is there.  There is a non-zero chance that 
data may be lost due to a panic in this instance, but the only 
alternative would be to dig out the backup tapes and do a full 
restore (guaranteeing loss of whatever data was generated since the 
last backup...).

Later...

Greg Oster