Conference smurf::ase

    Dave,
    
    Before someone claims this is or is not supported you will probably be
    asked to provide a more detailed picture.  Including scsi devices,
    cable part numbers, termination, cable lengths, etc.
    
    Regards,
    
    -- Greg

    Auh Greg, I knew somebody would want more info.  :-)
    
    The system (ASE) is not yet configured.  What the customer has now is
    two 4100s (not in an ase).  Each 4100 has 5 SCSI adapters.  There is
    some EMC storage hanging off these SCSI busses.  At this point I don't
    know what adapters they have or what is in the EMC storage cabinet.
    I also don't know anything about the lengths of the SCSI cables.
    They want to buy an 8400 with 10 ase compliant SCSI adapters.  I know
    all that stuff is important but at this point I just want to know if
    I can do it, assuming all the parts are "ase compliant".  If the answer
    is I can't create an ASE this way or I can't use and 8400 with two
    4100s.  That kind of thing?  In particular, I was concerned that the
    ASE has three members but NOT one SCSI bus common to all three.  It
    seems like it should work to me but I don't know what goes on behind
    the ASE curtain.
    
    Sorry to be so vague about the details but this is all the information I
    got and I don't want to go back for more information just to tell them
    that you can't have three members in an ase without having a common
    SCSI bus.  Or that the 8400 can not be in an ase with 2 4100s.  The
    bottom line is, If we support this, I'll make sure it is with the
    understanding that all the pieces and the whole must meet ase
    specifications.
    
    I guess I should have said this in the base note.  :-)
    
    
    Thanks,
    
    Dave S                                             

Actually, you are trying to build a configuration that is
intentionally broken.  Consider the following:

4100           8400           4100
scsi0----------scsi0---\  \---scsi0

scsi1---\  \---scsi1----------scsi1

This is a fully interconnected 3 node ASE configuration
that currently has 2 broken scsi buses.  It should be
supported when configured as such, and it should still be
supported when the broken cables happen.  The question, is;
is it supported when "permanently broken" scsi cables are
part of the configuration?  Or to put it another way, is it
supported when the 2 scsi cables are permanently NOT part
of the configuration?

The drivers don't care, so I'll leave the "support" question
to the ASE and product management types.

	Fred

    re .3
    
    Hi Fred,  Yes, that was my thinking too.  But I wouldn't have put it
    as well as you did.
    
    Dave S

    Hi,
    
    I think that you need AT LEAST ONE common SCSI bus to simply be able to
    set your ASE cluster. Without this bus, you will have to create the
    cluster from the 8400 (add member to). Because the lowest numbered ASE
    bus from the 8400 will NOT see each member, I will have to "ping" the
    not seen member on the second SCSI, and I think that your daemon.log
    will record that error every time.
    
    So, if you have at least one common SCSI bus (which must be the lowest
    numbered SCSI in each of the three systems), you will be in a three
    member ASE cluster with the oarticular situation that you have 5 scsi
    buses broken between  one of the 4100 and the 8400 and also 5 SCSI buses
    broken between the other 4100 and the 8400.
    
    Thus, for me, it can work with one common SCSI, but will be most likely
    NOT supported. Do'nt forget to use "restricted to favor member"
    placement policcy.
    
    There is another note speaking about that type of configuration, but I
    was not able to find it ???
    
    Manu.
    

    Thanks Manu,
    
    Oh yes, I won't forget the "restricted to favor member" placement
    policy.  :-)
    
    I agree that management of the ASE could become difficult if there
    is a problem with the 8400 is this case.  But as Fred said in .3,
    this is really a configuration that is built like it has all common
    SCSI busses broken.  I beleive the ASE will function properly
    if it gets to this condition and I think I can build the ASE (from the
    8400).  If I can't, somebody hit me on the head now because I have not
    done this before.  But the question is, if this customer has problems
    running an ASE in this configuration will they get support all the way
    up/down the Digital food chain?
    
    
    The problem is that the 4100s have no room for a 6th KZPSA and the
    customer feels that they need all the I/O preformance they are getting
    on each 4100, from the 5 SCSI busses.  They can not give up a SCSI bus.
    The customer has it in their mind that they want an 8400 and the rep
    would like to sell them one.  If this won't work (won't be supported)
    then the rep will try to sell them two 4100s and create two ASEs I
    guess.
    
    I'm not sure who can give the official position on the support question
    but I'd like to get this settled so that the rep can close this sale
    in Q4.  I just want to make sure there is enough time in case they have
    to go back with the two 4100s as a solution.  I don't know how much
    time that will take but I do know the customer is ready to buy the
    8400 if Digital will support the ASE configuration.
    
    
    Thanks,
    
    Dave S

...this is documented as being an invalid configuration.  This has
been discussed before in various notes.  Note 1627.2 has the quote
from the manual.  I really doubt the code will form a cluster when
booted using this config.

Dave
[Posted by WWW Notes gateway]

    Thanks Dave,
    
    I looked around for other notes but didn't come across 1627.
    I also tried to look through the docs but didn't have any idea
    which one might have the info.  I didn't find it where I was
    looking. :-(
    
    Dave S