[plug] Stressing a disk

[James Devenish]

On Thu, Jan 09, 2003 at 11:42:22AM +0800, Trevor Phillips wrote:
> On Thursday 09 January 2003 11:31, Cameron Patrick wrote:
> > Um, won't the md5sums produced be meaningless, as they are computed at
> > the same time as the disk writes?  Plus you could end up with several
> > dd's running after a few iterations if they take longer to complete than
> > the md5sum.  I'd not want to try it even on a disc without a live
> > filesystem :)
> 
> I think the point of the md5sum was to be another non-backgrounded read of the 
> partition. Theoretically it should take longer to md5sum than the dd's, thus 
> preventing multiple iterations of the backgrounded dd's.

OTOH if the original idea was that you did the dd's and md5s
sequentially and expected the md5s to stay the same (i.e. almost any
silent/corrected errors would produce a tell-tale change in the md5s)
then you would lose this benefit unless the backgrounded writes
converged.