> The Overland storage was not completing command in the specified
> time, and as a result Initiator would time out the commands and cleanup
> its queues. The command would complete at some point later and
> Initiator would flag it is "unknown itt" since it has already flushed the
> queues. The timeouts would certainly degrade the performance.

  In Pierre's posted logs there are not timeouts or protocol errors.  The dtrace
points to the target cutting the connection not the other way around.  I have
only taken a quick peek at the wire trace and again the target is cutting the
connection.  This is probably a different issue than the one listed above.

  Nigel you looking at the right time frames.  I'm not convinced yet the task
management response is related.  Its not consistent with the other disconnects
in the trace.

  My initial gut reaction from looking at the traces is there are two problems.
One is a possible configuration issue and the second is a target or initiator 
bug.
As for the first problem your getting way too many TCP retransmissions from the 
target to initiator also the transmission window is closing frequently.  This 
in itself is a performance problem.  In the first trace are there any 10 or 100 
Mb hops in the network between the target and initiator?
 
 
This message posted from opensolaris.org
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss

Reply via email to