spork at bway.net
Mon Sep 20 15:38:52 PDT 2004
On Thu, 16 Sep 2004, Justin T. Gibbs wrote:
>> Looking more closely, those backplanes should probably be chained:
> They must be chained if you want the system to behave correctly. The
> maximum parallel SCSI stub length is ~2in. which I'm sure is shorter
> than your backplanes.
Interesting. Saturday we went in to replace all 4 18GB drives with 36GB
models. While I was in the box, I recabled so that the chain went in one
backplane, and then to the other with a seperate cable. Also put a
terminator on the second backplane's "end" of the chain. With the
terminator on, no drives were found by the controller. With the
terminator removed, 4 drives showed up, but one would always show a sync
rate of 80MB/s vs 160MB/s on the others.
So something is obviously wrong. We ended up just using the one
backplane, mirroring two drives and putting a third in as a hot spare. I
wonder if perhaps the whole mess is not new enough to support UW SCSI @
Last question... Are the layouts of these SCA backplane cards at all
standardized? I'd gladly swap these out for anything that included docs
and is known to work when chained together. We can find no identification
on these things. This is a 2U box; each backplane stacks three drives and
is secured to the drive cages with a screw on each corner.
More information about the freebsd-scsi