[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ssag::ask_ssag

Title:Ask the Storage Architecture Group
Notice:Check out our web page at http://www-starch.shr.dec.com
Moderator:SSAG::TERZAN
Created:Wed Oct 15 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:6756
Total number of notes:25276

6402.0. "SDQUAL 0, BUFMODE 0(X), target shall not report good status." by GIDDAY::BENBROOK (Brisbane ruLes) Mon Feb 17 1997 19:47

	  I'm willing to bet my next pay on me - but please answer this for
	me (and quote book references too, no matter which way it goes).

	  Apart from Note 5827, I've seen no reference to this.


	  I have a customer that has taken it into his head that the SDQUAL
	field in the OVMS errorlog is telling him that his drives have an
	error in that they are saying "I'm done" BEFORE they really are.

	  The errorlog has words that say :-

		SDQUAL		00	BUFMODE = 0(X)
					Target shall not respond GOOD status
					on write cmds until data is written

	  He reads this as "Target SHOULD NOT respond .... " and assumes they
	are acting incorrectly.    I read it to mean that (because the HSJ40
	has Read cache only) the data will be written completely before
	the command is deemed complete (unlike WriteBack cache for example).

	  Of course, the only time he sees this is in an Errorlog entry - and,
	if I am correct, he will see it for EVERY errorlog entry.   I believe
	he is therefore thinking of MASSIVE POTENTIAL DATA CORRUPTION and 
	is concerned that "These RZ28D's are not very good !!!"

	  Does he have a case ?    Or is this an Informational entry ONLY,
	and it is simply saying "read cache enabled" ?

Les Benbrook,

MCS, Brisbane
T.RTitleUserPersonal
Name		DateLines
6402.1hmmm...SUBSYS::VIDIOT::PATENAUDEAsk your boss for ARRAY's...Tue Feb 18 1997 07:3925
I do not know what the author intended that to mean, but that verbage has been
the same as far back as I can remember.

 ERROR SEQUENCE 24.                              LOGGED ON:        SID 17000202
 DATE/TIME 24-SEP-1995 11:00:52.37                            SYS_TYPE 01400001
 SYSTEM UPTIME: 0 DAYS 00:51:41
 SCS NODE: RDOPS8                                              VAX/VMS V5.5-2

  '''''  bla bla bla ''''

       SDQUAL                00
                                       BUFMODE = 0(X)
                                       Target shall not report GOOD status -
                                       on write cmds until data is written

Here is an old log off an HSJ40/RZ28 (Not RZ28M), no WBC, no raid, no nothing.
At the drive level we do NOT have any drives with WRITE cache enabled and none
of our drives will report GOOD status until after the data is committed to the
media. I cannot address what controllers do between them and the hosts.

Roger.
6402.2A controller to Drive I/O option.SSDEVO::MARTENSQualification Program ManagerTue Feb 18 1997 08:0237
    
    Okay, I will take a stab at this, 
    
    The decode of the error log entry can be a bit misleading, while the
    message is correct, without the full context a person may not
    understand what other options may exist.
    
    As I understand the field, 
    
    The SCSI buffered modes are as follows
    
    0 = The target shall not report GOOD status on write commands until
    the data blocks are actually written on the medium.
    
    1 = The target may report GOOD status on write commands as soon as all
    the data specified in the write command have been transfered to the 
    targets buffer. One or more blocks may be buffered prior to writing 
    the block(s) to the medium.
    
    2 = The target may report GOOD status on write commands as soon as:
    (1) All the data specified in the write command has been successfully
    transfered to the target's buffer, and (2) All the buffered data from 
    different initiators has been successfully written to the medium.
    
    
    This field is related to the controller communicating with the drive,
    and what type of I/O buffering is requested. This I belive is the
    default, since we always request that the drive does not use a
    writeback cache.
    
    So, the error log entry is telling you that the I/O SCSI buffered mode
    was set to ensure that the data was written to the media, BEFORE the
    drive could send back good status.
    
    Does this help?
    Bert
6402.3Nice to have it resolved - thanksGIDDAY::BENBROOKBrisbane ruLesTue Feb 18 1997 16:4111
	  Those answers are just what I wanted to see (I really did not
	want/expect to lose a month's pay     ;^)


	  So, in short, it is Informational and NOT indicative of an error.

	Thanks to both respondents,

Les
6402.4SUBSYS::VIDIOT::PATENAUDEAsk your boss for ARRAY's...Wed Feb 19 1997 08:242
Correct. It's indicitive the the type of I/O performed.