In article <20030904200523.GB13820@moghedien.mukappabeta.net>,
	Matthias Buelow <mkb@mukappabeta.de> writes:
> * Instead of panicking, or failing/hanging processes that could not
>   write to the disk (even sync(1m) didn't hang), it rerouted writes for
>   the failed device to some other data store

No, it probably just kept the dirty pages in the buffer cache. This also
explains the bad performance. If the system needs memory it will steal
pages from the buffer cache. It can of course only do that for pages
which haven't changed or have been written back to the storage. As it
couldn't write dirty pages back to your disk it had to get them by
paging out normal memory.

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/