SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    Re: iSCSI: LUs ownership/iSCSI MIB



    
    Pierre,
    
    I have some comments on your note with respect to SAM/SCSI.
    
    1) The interpretation depends a lot on what one means by "target".  The
    default notion of this term is Target Port (not device, though that is not
    necessarily well captured by the MIB).   Clearly SAM-2 allows multiple
    target ports to be i/o paths to a common logical unit.
    2) the "asymmetric port" proposal from Ken Moe of Sun in T10 (for SPC-3) is
    designed to deal with SCSI's view of this active/passive picture you
    describe.
    3) I don't believe I've ever seen anything in SAM that precludes even
    multiple Target Devices from sharing a logical unit.  I believe the
    architecture is silent on this, which means that it's OK.
    4) Also what's "viewable" by the MIB interface may NOT be the same as
    viewable by Report LUNS.  The first is a management interface to get a view
    of the whole target device.  The other is a specific host's (more
    precisely, initiator's) view of the accessible logical units on a
    particular port.  So, the MIB could easily present the same answer on all
    ports *of the same target device* and a different target device may present
    a different answer.
    
    All this confusion about multiple pathing, etc. is one reason for logical
    units to have unique identifiers (EVPD page83h) AND (in my opinion) to keep
    most of this discussion out of iSCSI MIB.
    
    Finally, some of us in the N&DT team are wrestling with the mapping of SAM
    constructs to iSCSI constructs.  The resolution of that may clarify the
    ambiguity of the term "target" in the MIB.
    
    Jim Hafner
    
    
    Pierre Labat <pierre_labat@hp.com>@ece.cmu.edu on 03-23-2001 10:34:56 AM
    
    Sent by:  owner-ips@ece.cmu.edu
    
    
    To:   ips@ece.cmu.edu
    cc:
    Subject:  iSCSI: LUs ownership/iSCSI MIB
    
    
    
    Mark (bakke),
    
    
    Where
    =====
    In the MIB draft, in the chapter
    
    5.1.  Overall MIB structure
    
    
    There is:
    
    "Therefore, LUs are "owned" by Targets, and LUNs are owned
       by Sessions."
    
    Problem
    ======
    
    If a LU is owned by only one target how do you deal with an
    active-passive
    configuration such as:
    active link: portal IP1, LUN=1
    passive: portal IP2, LUN=2 (but it is the same LU as the one accessed
    through the active link)
    
    As the various portals of a target provide the same LUN view, IP1 and
    IP2 can NOT
    be two portals of the same target.
    
    In this case we would need two target names, one for the active link and
    one for the passive.
    
    But if a LU can be owned by only one target we can't do it.
    I think SAM doesn't allows a LU to be shared by two targets and that
    seems
    to be reflected in the chapter "5.1.  Overall MIB structure".
    
    Solution
    ======
    In the case of an active-passive configuration, the target must
    guarantee the same
    view on the two paths (active and passive).
    Is it a problem for whose who build such targets?
    
    
    Regards,
    
    Pierre
    
    
    
    
    
    
    
    


Home

Last updated: Tue Sep 04 01:05:16 2001
6315 messages in chronological order