[nycbug-talk] tape drive woes

Dan Langille dan
Fri Feb 4 11:19:37 EST 2005

On 3 Feb 2005 at 22:11, Tillman Hodgson wrote:

> The traditional wisdom with bus resets is that it's due to exceeding
> the SCSI specifications in some manner: poor termination, too-long
> cable, that sort of thing. I've have luck with a DEC StorageWorks JBOD
> tower that used to bus reset like a mad banshee simply by telling the
> controller to run it at a slower rate -- a more robust SCSI bus
> results from the "easier" specifications. Since this particular SCSI
> controller only "auto-detects" SCSI settings and won't allow me
> override them in it's "BIOS", I wrote a /etc/rc.d script that manually
> forces the devices down one speed notch just before fsck kicks in
> (which would cause bus resets like crazy). The basic idea is this:
>   camcontrol negotiate da0 -R10 -q -a
> I'm simply moving the bus from 20MHz to 10MHz which, at 16 bits wide,
> moves me from 40MB/s to 20MB/s. That's more than fast enough for the
> drives I have in the JBOD (which aren't ever accessed simultaneously
> with the use I've put them to) and all top out at about 12MB/s
> anyway).

I've not tried that yet.  Last night, I tried a DDS-2 tape. I was 
able to back up.  But I still think something is wrong.  It's a DDS-4 
drive, and I was unable to do a backup on a DDS-4 tape.

While it was labelling the tape, a bit of tape would be fed through, 
then the unit would "click" three times, rewind, and read the tape 
again, click three times, repeat... several times.
Dan Langille : http://www.langille.org/
BSDCan - The Technical BSD Conference - http://www.bsdcan.org/

More information about the talk mailing list