Thanks Matt and Jon (and later posters..)

> 
> AFAIK, the policy is "if I don't step on your feet, and you 
> don't step on
> mine, everything will work okay."

Cool!  I presume also that standard ANSI sequences are used so
that disimilar cards won't be much of a problem?  

The only experience I had with this was on Sequent Clusters 
(which used cards that were the same model and firmware) and they 
used their DLM (distributed lock manager) to regulate access to 
certain areas of the disk, but it was posible to for instance mount 
sd0a on one machine and sd0e on another machine, even though they 
were on the same disk.  I don't think they were locking out the whole but
on a slice (or partition) level, where with tapes it was the 
entire device locked out.

> they _never_ try to use it at the same time... I remember having both
> access it at the same time, and the result was a bunch of 
> SCSI errors and one hung machine :-(

At which level would I expect to have problems with disks, or is it
as was said earlier about not stepping on each other.  Sorry to keep
dogging this question, but the answer was a little ambiguous.  I don't
know if I shouldn't have two nodes accesing the same disk because: 

	A.  There will be disk errors.

	B.  Because the filesystem will be toast...

      (OR C)  Both.

If the answer is 'B' then it is rather academic to regulate mounts
of a filesystem, but 'A' would be a little more tougher.  'C' I
don't care about because of 'A'.


> 
> It would be neat to be able to have a machine mount some SCSI disks as
> r/w, then have a few other machines mount them r/o, but I 
> think there are caching issues with this right now.> 

Well we just need to find the 'Jason Thorpe' of the network drivers
for filesystems and then it too would probably be rather academic :-)
(I hope this is taken as a compliment)...

Thanks again...



Andy