SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    Re: Asynchronous Message: Session Termination



    
    Santosh-
    
    I agree with all of your proposed modifications; that's what
    I had in mind as well.  Codes 3 and 4 will then do a better job
    of code 2 (target requests logout), so 2 can be removed.
    
    I also think that we should be more specific on
    "... will drop all connections" and say
    "... will drop all connections within the current session".
    I think that was implied, but it is probably better to
    explicitly state it.
    
    I agree with both you and Julian that the statement about initiators
    modifying these time values should be removed; I think that these
    time values are really up to the target, and we should remove
    them from the negotiation.
    
    --
    Mark
    
    Santosh Rao wrote:
    > 
    > Julian,
    > 
    > The modified text on "Async Message" seems to have missed out several
    > aspects of the discussion on this thread :
    > 
    > 1) It is useful to provide a parameter for the "target will drop this
    > connection" and "target will drop all connections" to indicate the time,
    > in seconds, after which the target would drop the connection[s].
    > 
    > This allows :
    > 
    > - the target to warn the initiator ahead of time about an upcoming
    > disruption so that initiator can take action accordingly.
    > 
    > - Allows elimination of the "target requests logout" option since the
    > "will drop connection/all connections" flavor provides the initiator an
    > upper time bound after which the connection[s] are going to be dropped.
    > This allows the initiator to attempt to quiesce the connection[s] prior
    > to the expiry of that time period. The "target requests logout" option
    > should be removed, if its equivalent functionality or better can be
    > provided by the "will drop connection/all connections" flavors.
    > 
    > 2) Suggest modification of the below wording :
    > " If the initiator does not attempt to reconnect within the time
    > specified by Parameter 3 or Parameter 3 is 0 the session is terminated -
    > in which case no other responses should be expected from the target for
    > outstanding commands on this session and the initiator should terminate
    > them appropriately."
    > 
    > to indicate that upon expiry of the LogoutLoginMaxTime (parameter 3),
    > the target internally aborts all outstanding commands on that connection
    > and the initiator should consider those I/Os implicitly aborted. No
    > explicit Abort mechanisms are required from the initiator to the target
    > upon expiry of LogoutLoginMaxTime to terminate these commands.
    > 
    > 3) The "target will drop all connections" flavor description should
    > indicate that the target is performing a session logout. I like the
    > parameter descriptions provided by Mark [Bakke] for this flavor, namely
    > :
    > 
    > - Parameter2 indicates the time, in seconds, after which the target will
    > drop all connections.
    > - Parameter3 indicates the additional time, in seconds, after the expiry
    > of Parameter2, that the initiator must wait before attempting to
    > re-establish a session.
    > 
    > The LogoutLoginMaxTime in the current defn of "target will drop all
    > connections" (Parameter 3) provides little value, since the target
    > should be aborting all oustanding I/O as a part of that session logout
    > and not wait until LogoutLoginMaxTime.
    > 
    > 4) Regarding :
    > "Initiators can attempt to set/modify the values a target will use for
    > Parameter2 and Parameter3. However - I am unclear if we should leave the
    > initiator the degree of control implied by (innocuous looking) last
    > statement."
    > 
    > Since, LogoutLoginMaxTime & LogoutLoginMinTime are only applicable in
    > the context of a target originated "connection drop" as indicated by the
    > async events, these keys add little or no value in the login
    > negotiation. In all cases, it is the time values provided by the target
    > in the async events that are applicable and over-ride any login
    > negotiated values for these keys. Hence, should these keys be removed
    > from the login negotiation ?
    > 
    > Regards,
    > Santosh
    > 
    > julian_satran@il.ibm.com wrote:
    > >
    > > Mathhew, Mark, Matt & Santosh:
    > >
    > > Here is the new text for 2.17.1
    > >
    > > 1.1.1     iSCSI Event
    > >
    > >    The codes used for iSCSI Asynchronous Messages (Events) are:
    > >
    > >       0    No iSCSI Event
    > >       1    Target is being reset.
    > >       2    Target requests Logout - the Parameter1 field indicates on what
    > >       CID.
    > >       3    Target indicates it will drop the connection - the Parameter1
    > >       field will indicate on what CID while the Parameter2 field indicates,
    > >       in seconds, the minimum time to wait before attempting to reconnect
    > >       and Parameter3 the maximum time to reconnect  after the initial wait
    > >       (Parameter2). If the initiator does not attempt to reconnect within
    > >       the time specified by Parameter 3 or Parameter 3 is 0 no other
    > >       responses should be expected from the target for outstanding commands
    > >       on this connection and the initiator should terminate them
    > >       appropriately. A value of 0 for Parameter 2 indicates that reconnect
    > >       can be attempted immediately.
    > >       4    Target indicates it will drop all connections - the Parameter2
    > >       field indicates, in seconds, the minimum time to wait before
    > >       attempting to reconnect and Parameter3 the maximum time to reconnect
    > >       after the initial wait (Parameter2). If the initiator does not
    > >       attempt to reconnect within the time specified by Parameter 3 or
    > >       Parameter 3 is 0 the session is terminated - in which case no other
    > >       responses should be expected from the target for outstanding commands
    > >       on this session and the initiator should terminate them
    > >       appropriately.  A value of 0 for Parameter 2 indicates that reconnect
    > >       can be attempted immediately.
    > >
    > >       Initiators can attempt to set/modify the values a target will use for
    > >       Parameter2 and Parameter3
    > >
    > >       However - I am unclear if we should leave the initiator the degree of
    > >       control implied by (innocuous looking) last statement.
    > >
    > >       Thanks for helping make this text better.
    > >
    > >       Julo
    
    -- 
    Mark A. Bakke
    Cisco Systems
    mbakke@cisco.com
    763.398.1054
    


Home

Last updated: Tue Sep 04 01:04:37 2001
6315 messages in chronological order