[tahoe-dev] Tahoe on large filesystems

[Brian Warner]

On 2/4/11 4:51 AM, Shawn Willden wrote:
> 
> Without getting into the advantages/disadvantages of that approach,
> the reason I mention it is because what I've observed is that when a
> disk gets an I/O error on any one of the partitions, the OS assumes
> that the whole disk is having trouble and drops _all_ the partitions
> out of their RAID arrays.

Is it the kernel's filesystem code that makes that decision, or the
kernel's block-device driver?

Here's a crazy idea: build a small block-allocation filesystem into the
Tahoe storage server (no more clever than, say, FAT16), and let Tahoe
write to the raw disk. If the kernel will give you EIO on one block, but
then let you keep reading and writing to others, Tahoe could return an
error for that one share, but keep serving and accepting other shares.

Probably not worth the effort. You'd need to build a share-lookup index
as well, which Mnet did, but was a hassle. Folks have put a lot of work
into modern filesystems, and you'd have to repeat at least some fraction
of that.

cheers,
 -Brian