[4878] in linux-scsi channel archive
Re: dual-attached disk
daemon@ATHENA.MIT.EDU (thoth@purplefrog.com)
Tue Oct 13 23:44:35 1998
From: thoth@purplefrog.com
To: linux-scsi@vger.rutgers.edu
In-reply-to: Your message of "Tue, 13 Oct 1998 12:10:01 PDT."
<3623A589.7EF4@nis.acs.uci.edu>
Date: Tue, 13 Oct 1998 15:38:04 -0400
Dan Stromberg <strombrg@nis.acs.uci.edu> ,in message <3623A589.7EF4@nis.acs.uci
.edu>, wrote:
> In the sun world, you don't "just" connect the cable - it requires extra
> software. Otherwise the buffer caches would get out of synch with what's
> on the disk and really mess things up.
I have never heard of anyone attempting simultaneous access to the disks. I
wouldn't be surprised if it were done in a research situation, but the
arbitration would be a pain in the ass, code-wise.
> Now I suppose if you can only have one system access the disk at a time
> - say the second only uses the disk if the first has certifiably
> crashed, then that might be another matter.
This is the standard model. One machine "locks" the disks, and the other
machine takes them over when it is sure the other is a corpse (or when ordered
to by the system operator, which is hopefully the normal case).
There are Sparcs operating in this configuration under software from Veritas
(and other vendors) in corporate machine rooms all over the country. Heck,
there's one in my roommate's spare bedroom.
As I understand it, the trick is to have the SCSI drivers on both ends "play
nice" and not battle for control of the bus. I saw this on a wish list for
the Linux kernel under the "clustering" heading. I wondered if anyone had
made any progress on it.
--
Bob Forsman thoth@gainesville.fl.us
http://www.gainesville.fl.us/~thoth/
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
