RE: iSCSI: Need to Kill Session from Surviving Node

["Y P Cheng" <ycheng@connectcom.net>]

> Wait a minute, I did not expect my Surviving Node question to somehow
> trigger the discussion of whether the Target Reset should be issued by
> iSCSI Initiators.   That is a completely different Pit.
>
> My question was, when the Storage Controller fails over to a surviving
> NODE, how do we tell the Initiator that is waiting peacefully for the
> TCP/IP connection to time-out.  To quickly give-up and let the initiator's
> SCSI layer, retry the operations, so that the iSCSI layer can start either
> sending the SCSI cmds to the Surviving Node, or starts a new Session with
> the Surviving Node.
>
> This has nothing to do with the Initiator issuing Target Resets.
> .
> John L. Hufferd

John,

I think the FC-PLDA is a very good document defining the recovery behavior
of the SCSI initiator and target for fibre channel.

In my experience of designing SCSI adapters, we never tried to do failover
on the target side and then inform the initiator to go to a different
connection.  This is because no matter how hard a target tries, it can't
detect a missing command lost in transit.  Ultimately, the failover must be
started by the initiator.

I don't agree on optimizing a recovery by speeding up the restart because if
connection failure happens often we have something much bigger to worry
about than how fast we do failover. If a connection fails, the initiator on
its own must detect the failure by either time out or the frequency of
failure exceeding certain threshold.  If a target could inform the initiator
about the troubled connection using another connection, I am not sure it
would worth the added complexity.  After all, if the target knows about the
problem, the initiator would know equally fast as well.

Check out the FC-PLDA and you will see what I meant.  Just my two cents.

Y.P.