[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference vaxaxp::vmsnotes

Title:VAX and Alpha VMS
Notice:This is a new VMSnotes, please read note 2.1
Moderator:VAXAXP::BERNARDO
Created:Wed Jan 22 1997
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:703
Total number of notes:3722

362.0. "same disk unit cluster on SCSI" by HGOSPS::MASCOTTANG () Fri Mar 21 1997 04:04

   My customer buys two ALPHA 4100 systems and runs ALPHA VMS V6.2. His 
configuration is listed as follows:

        ---------+--------------------------+----------------
                 |                          |
                 |                          |
 +-------+  +-----+-----+            +-------+-----+    +-------+
 | disk2 +--+  NODE A   |            |   NODE B    +----| disk1 |
 +-------+  +----+------+            +------+------+    +-------+
                |                          |
                |       +---------+        |
                +-------|  HSZ40  |--------+
                        +---------+

My question is :

1) Can I have the same unit number on disk1 and disk2 ?
2) If yes, how can I do that ?

   Would anyone have this experience ? Any advcie should be appreciated.

    
    
T.RTitleUserPersonal
Name
DateLines
362.1UTRTSC::utojvdbu1.uto.dec.com::JurVanDerBurgChange mode to Panic!Fri Mar 21 1997 07:2710
>1) Can I have the same unit number on disk1 and disk2 ?

Yes.

2) If yes, how can I do that ?

Sure, give each system a unique allocation class.

Jur.

362.2AMCFAC::RABAHYdtn 471-5160, outside 1-810-347-5160Fri Mar 21 1997 08:585
re .1:

>... give each system a unique allocation class.

Um, that'll be ill-advised given the shared devices.
362.3Use Unique Unit Numbers, Or Upgrade To V7.1XDELTA::HOFFMANSteve, OpenVMS EngineeringFri Mar 21 1997 10:0242
:   My customer buys two ALPHA 4100 systems and runs ALPHA VMS V6.2. His 
:configuration is listed as follows:

   Note: the customer *must* run V6.2-1H3, or V7.1.  V6.2 will *not* work.

   One must have unique unit numbers across all disks within any particular
   disk allocation class (zero or non-zero), and across all tapes within any
   particular tape allocation class (zero or non-zero).

   Shared storage interconnects normally require the use of the same
   non-zero allocation class and unique unit numbers, to uniquely and
   explicitly identify all direct and served paths to a particular device
   with a particular unit number as being different paths to the same
   device, and not as different devices with overlapping unit numbers.

   This particular configuration is one of the reason why OpenVMS Alpha
   V7.1 added port-level allocation classes.

   I'd recommend an upgrade to V7.1, or I'd split the SCSI bus, or I'd
   use unique device unit numbers.

   If the customer decides to not upgrade, and to use seperate non-zero
   allocation class numbers on the two Rawhide systems, they must still
   keep the unit numbers on each Rawhide unique, and they will loose the
   ability to fail-over served access to the shared SCSI between the two
   Rawhide systems. 

   If the customer chooses to use overlapping unit numbers and disjoint
   allocation classes, the customer must avoid mounting any disks on the
   shared SCSI that are mounted by the other Rawhide system.  Attempts
   to share disks are likely to result in severe data corruptions on
   the shared disks.

   I would not recommend this disjoint-allocation-class configuration.
   I would configure unique unit numbers, and would place both Rawhide
   systems into the same allocation class.  If I cannot change the disk
   unit numbers, I'd look to upgrade to OpenVMS Alpha V7.1, and use the
   port-level allocation class support.

   And SPEZKO::CLUSTER is a better spot for VMScluster questions...

362.4EVMS::KUEHNELAndy K�hnelFri Mar 21 1997 12:1519
    Just because Steve didn't really tell you how to use port allocation
    classes...
    
    The SCSI bus with the non-shared devices should get a port allocation
    of 0.  This turns the device name into "node$ddcu". for these devices
    _regardless_ of the system allocation class.
    
    If the shared SCSI bus has the same controller letter on both systems,
    you can give both systems the same allocation class.  If the controller
    letter differs, the bus must get the same port allocation class > 0 on
    both systems.  This will turn the device name into $n$dka... cluster
    wide.
    
    For V7.1, if you use port allocation classes at all on a system, the PKA
    bus must have a port allocation class, either zero or positive.  I am
    currently working on a fix that will allow to lift this restriction.
    
    
    -andy
362.5CSC32::M_DIFABIOMOVL #OPINION,EXE$GL_BLAKHOLETue Mar 25 1997 15:173
    Port allocation classes only exist for V7.1, correct?
    
                Mark d.
362.6V7.1 and later...XDELTA::HOFFMANSteve, OpenVMS EngineeringTue Mar 25 1997 16:014
:    Port allocation classes only exist for V7.1, correct?

   Port allocation class support exists in V7.1 and later.