FCP-2 public review comment resolution, changes to various documents

Robert Snively rsnively at Brocade.COM
Tue Sep 11 07:01:18 PDT 2001


* From the T10 Reflector (t10 at t10.org), posted by:
* Robert Snively <rsnively at brocade.com>
*
To:		T11, T10

From:  	Bob Snively

Date:		September 10, 2001

Subject:	Meeting resolving FCP-2 public review comment

INTRODUCTION:

A public review comment was received on FCP-2.  The
comment indicated that early implementations of
FCP-2 tape environments were unable to continue
normal operation after certain induced errors.  Review
of these recovery failures is still in progress, but
appears to be related to the execution of PLOGI
during the recovery process.  The PLOGI resets the
tape drive's mode pages, making normal continuation
impossible.  

The resolutions below summarize the changes in various
standards required to assure proper continuation of
tape operations in the presence of certain types of
errors. 

RESOLUTION:

FC-FS

	The document will recommend that PLOGI not be used
	unless re-authentication after a temporary loss
	of a link is unsuccessful.

FC-MI

	The document will require that PLOGI not be used
	unless re-authentication after a temporary loss
	of a link is unsuccessful.

FC-DA

	The document will require that PLOGI not be
	used unless re-authentication after a temporary loss
	of a link is unsuccessful.

SSC-2

	The document will provide requirements on
	mode pages and other mode select parameters that
	will allow continued normal operation after a
	reset.  Recommendations will be made about the
	behavior of mode pages and mode select parameters.
	The recommendations will encourage saveability.
	In addition, the definition of current and default
	parameters will allow for the properties of the
	tape that is loaded. 

Note that some implementations of tape drives and host
adapters may be affected by these modifications.

No changes are required for FCP-2.

DISCUSSION AT THE MEETING:

1)  Testing status

	Roger Cummings presented testing status on the
	study of the FCP-2 ballot problem.

	The good news:

		PLOGI causes backup jobs to fail by 
			resetting mode pages in mid-media.
		Unplug and replug also mostly kills the backup.

	The bad news:

		We know almost nothing new.

		None of the 3 recently received FC tape
		drives, but none of them are ready for
		prime time.  None have yielded results
		in which they have confidence.  None will
		support their standard acceptance test.

	New test case:

		Unplug receive side of tape drive only.

		In point-to-point

		   4 out of 4 HBAs produce PLOGI at some
		   point in the loop reinitialization process.

		Fabric

		   Tape drive issues successful FLOGI and
		   recovers.

	The situation is frustrating.

	They need more FCP-2 compliant equipment (HBAs and
		tape drives)

	Recommendations:

		End consideration of unplug/replug case.

		Accept Charles Binford's suggestion from
		last time of a bit in Process LOGI to
		select FCP behavior (i.e. no reset actions
		from link-level events).

		Add a note to FC-FS about the dangers of
		PLOGI during "normal" fabric operation,
		proposing the use of PDISC instead.

2)  Discussion:
	
	Dave Peterson:

		Suggests the problem is HBA configuration
		verification failures.

	There was general agreement that HBAs should
	not be performing PLOGI under the conditions that
	have been identified.  FC-PLDA, FCP-2, and other
	documents indicate that authentication should not
	use PLOGI.

	However, because there is an installed base of
	HBAs that do perform PLOGI, previous tape drives
	have explicitly ignored PLOGI with respect to
	resetting the mode pages and in fact leave the
	state of the media and mode pages unchanged.

	Those more up-to-date tape drives that correctly
	implement the resetting of the mode pages are
	caught by the legacy HBAs' behavior.

3)  Proposed solutions from the committee:

	FC-FS and FC-MI shall be changed.  FC-FS will 
	recommend that PLOGI not be used unless 
	re-authentication is unsuccessful.   FC-MI and FC-DA will
	require that PLOGI not be used unless re-authentication
	is unsuccessful.  That will result in this
	restriction being tested by SAN Mark tests.

	Saved mode pages would be required.  Requires 
	changes to both the application (to set the
	save bit) and to the drives (to provide saved
	pages).  Some of these may also reflect into
	some driver stacks.  To find the parameters automatically
	without using saved mode parameters may require
	examining the tape at BOT.  This would go in
	SSC-2 and maybe in FC-DA.

	The tape drive's "default parameters" should probably
	reflect the actual present media installed.  This
	requires changes to the present tape and SSC-2.
	If BOT access is required, this is a multi-minute
	impact on the back-up process.  The preferred
	mechanism is to use saved parameters.  

	It is really key that parameters that can only be
	established at BOT be remembered one way or another.
	Later parameters (block length and others) 
	may be lost at the time of a reset.  Notification
	is provided by Unit Attention.  Note that this is
	still not much help, compared with saved parameters.
	This makes recovery vendor specific.

	Note that SSC-2 will have to include text that clarifies
	that the block descriptor and the device specific parameter
	(which SPC-2 defines as always "current") really means
	that the value at the time of the reset must be
	maintained.

	Note that different mode sense pages may require special
	reminders, especially with respect to partition and
	device configuration pages.

	Some tape vendors may be unhappy with this.  Applications
	may have to be tolerant of (if not happy about) devices
	that cannot save pages.

	It may be acceptable to create a "saveable" that is
	saveable across resets, but not power-down.

4)  Action items:

	All participants to review the proposed approaches.

	Robert Snively to prepare draft of public review 
	comment response for review by all participants.

	Roger Cummings accepted an action item to create
	appropriate text for FC-MI, FC-FS.

	Dave Peterson accepted an action item to include the
	proper concepts in FC-DA and in SSC-2.

ATTENDEES:

	Robert Snively
	Roger Cummings
	Ken Moe
	Kevin Butt
	Bob Nixon
	Dave Peterson
	Rob Elliott
	Ralph Weber
	Erich Oetting
	Paul Suhler

Bob Snively                        e-mail:    rsnively at brocade.com
Brocade Communications Systems     phone:  408 487 8135
1745 Technology Drive
San Jose, CA 95110

*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list