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

Re: Calling Customer Service for bad sectors



    Date: Sun, 20 Nov 88 18:56:25 CST
    From: forbus@p.cs.uiuc.edu (Kenneth Forbus)

	    Most of these utilities are avaiable.  The FEP provides a Disk Format
	    command. Each machine comes with an Initial File System (IFS) tape that
	    you can use to create a standard FEP directory structure after
	    reformatting.  The tools in sys:n-fep;fix-fep-block.lisp can splice out
	    bad blocks and generally repair things.  Of course, there is not much
	    documentation for the do-it-yourself disk repairer, but hackers can make
	    do.

    Well, the only thing I have in sys:n-fep; is a bunch of flod files.
    So I don't seem to have these tools.  
The formatting tool is in v127-disks.flod (or g20x-disks.flod if you have a G-machine).
					  This happened to me once, and I
    was told by Symbolics Software Support that in fact Release 6.1 does
    not have these tools.  
It is true that it didn't have SI:FIX-FEP-BLOCK; there was a version that
ran in 6.1 that was buggy.  The version released in 7.2 is considerably
cleaned up; that's why we decided to release it.
			   Someone sent me a function that did the trick,
    but only after hours of negotiation.
The lisp machine disk environment is quite a bit more complicated than that of PCs
(and perhaps even UNIX workstations); it is not clear whether it is a benefit
or a hindrance to customers to provide them with tools which, when used inappropriately,
might make things much worse.  The 6.1 tool once was used unthinkingly by someone
at Symbolics on a file server; it required approximately 20 hours of work by
some 2 or 3 people who all had better things to do to fix up the resulting mess.
At a customer site where the requisite knowledge probably would have been unavailable,
the time (and cost) would have been even greater.  It is probably this type of incident
which made whoever it was reluctant to release the tool

	    My point is that it is probably quicker, less painful, and quite
	    possibly cheaper for Mr. Srouji to have customer service take care of
	    the problem.  If he is on contract and he has recent LMFS backup tapes,
	    why suffer needlessly?

    Have you ever restored a large LMFS?  
Yes.
					  Doing the sector fix takes about
    20 seconds if you have the code.  
I don't know what you mean by "doing the sector fix".  How to fix a disk
problem depends upon what is wrong.  I can imagine simple things being
fixed in 20 seconds, but it would certainly take me much longer than 20
seconds to decide if the problem was simple enough that such an approach
would help more than hurt.
				      Compared to hours for restoring the
    LMFS, not counting down time waiting for customer service.
But what about the hours lost due to corrupted files, missing data, etc.
if you are wrong about the 20 second fix?

    And if you try to do board swaps, like you can with Sun, TI, and other
    vendors, you are in real trouble.  Over a month turn around time.  (TI
    gives universities five business days).  Worse yet, any accompanying
    documentation (such as what the problem is) is thrown away without
    examination.  Our technicians carefully isolated a nasty bug in a fep
    board that takes three days to show up when you swap it into a new
    machine.  When they sent the board in they included a letter which
    explained their procedure carefully.  A month later, the same board
    came back, with a note that said they had run the board for 24 hours
    and couldn't find a problem!  
This1 is0 distressing, it has happened to me, and I was similarly annoyed.
However, in my opinion, calling or writing Symbolics management in
a position to do something about it would be a better way to improve this
situation than complaining at/to SLUG.
				  We called, and found out that the
    standard procedure is to simply throw everything except for the board
    away without examination.  It has been several weeks since the board
    was sent back, without a sign of its return.  At least we received a
    verbal promise by a Symbolics engineer that they would indeed look at
    accompanying documentation this time.

    Do I sound annoyed?  Yep.  I love my Symbolics machines for the
    quality of their environment.  I used to love them for their
    performance.  I have never been impressed by Symbolics customer
    support, sales, or maintenance.  In fact, Symbolics is worse than any
    other vendor I know in these areas.  I sure wish it were otherwise.
    My students and I are tired of always having at least one machine out
    of six down waiting, waiting, waiting, for boards to come back.