[plug] handling failed non-redundant storage in a server

[Craig Ringer]

On Thu, 2004-02-12 at 13:05, Brad Campbell wrote:

> I found for a quick and dodgy solution while having temporarily connected high speed drives.
> I prop the drive up with a matchbox at each end (electrical insulation while providing a void under 
> the drive) and sit an 80mm low speed fan pointing at it. This keeps the disk shell to within about 
> 10 degrees of ambient under a full sustained load.

Interesting, if kind of scary. I'm going to see if I can find a cage
(with cooling) for SATA drives first, because it'd be nice to be able to
avoid having to plug in and unplug the drive each time I want to swap it
anyway.

Given that SATA drives appear to have identical power and data connector
positions, all that'd really be needed would be a "drive socket" - a 5
1/4 bay that you could slide a drive into, with fans for airflow and
connectors positioned properly at the back. No need for the old PATA or
SCSI-style drive cages. I wonder if something like that is available...

> These new Maxtor Maxline-II SATA drives are only detected as UDMA-33 under the stock kernel HPT366 
> driver (HPT374 card) and I can get at most about 15Mb/s to and from them.
> Using the Highpoint driver module compiled from source without the binary only raid module I can get 
> full speed to and from the drive. (About 20Mb/s sustained write and 50Mb/s read)

I'm seeing the latter numbers here - quite decent performance. Hence, in
part, the overheating. I'm using Maxline Plus II SATA (7200rpm) drives,
attached to a 3ware 8500-8 controller. It was, in fact, one of those
that caused this little incident.

Craig Ringer