Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!columbia!rutgers!husc6!cmcl2!brl-adm!adm!rackow@anl-mcs.arpa
From: rackow@anl-mcs.arpa (Gene Rackow)
Newsgroups: comp.unix.wizards
Subject: Re: BSD + CDC + Emulex
Message-ID: <7424@brl-adm.ARPA>
Date: Mon, 18-May-87 09:08:50 EDT
Article-I.D.: brl-adm.7424
Posted: Mon May 18 09:08:50 1987
Date-Received: Tue, 19-May-87 01:47:47 EDT
Sender: news@brl-adm.ARPA
Lines: 75

I hope CDC is listening an I'll get this problem resolved, I hate spending
the midnight hours formatting the soon to be powder.  I can't afford to
replace them with something better (read reliable).

Short summary first.  Use any drive with a CDC label as a boat anchor only.
I have had no problems with the Emulex controller.

System setup:  Vax 11/780 4.2bsd.  Drives are used for user file space
and 3rd and 4th swapping drives.  Other drives 2/rm05 (root;swap;usr;user),
1/rp06 (tmp;user), 2/rp07(user) 2/9771 (user;swap).  Clean "computer-room"
enviroment with several other machines and lots of disks.  In the years we
have had the VAX system, the only OTHER disk problems we have had are 1 crash on
an rm05 and 1 failing rp07, each after several years in use.  (Side note/Lots
of Eagles with NO problems)  (Also note other CDC exp. on different systems
here.)

FLAME ON | AFTERBURNER | NOVA | ...    (Lots and Lots of heat)

Time to do some yelling about REAL trash ----

We have been using the CDC 9771 drives for about 4 years.  The longest time
one has gone without a HEAD-CRASH is 16 months.  The average is about 10 months.
The last one was barely 4 months old.  You get very little if any warning
that anything is going wrong.  A soft error occurs about 10-12 hours before
crash, a hard error about 6-8 hours before, and then death.  We have had CDC
in to repair the drives time and time again.  Each time they claim we are the
only site with this kind of trouble, but we are about the only place using
them that they service from this office.

Some interesting notes:
1) For a drive that "has been out and stable" for some time, why is it that
every replacement head disk assembly (HDA) looks different.  A new filter
here, an added bracket there, remove the filter, etc.

2) When the service person arrives and checks the serial number of the HDA,
it is a "common" claim of "Oh! That is within serial number range ??-??,
where we recognize a problem of crashing heads.  This has been fixed
recently and you shouldn't have any more problems."   As they are leaving
I say "Bet I'll see you within the year to replace it." (Quotes not exact
as this is usually happening after many hours are spent rearranging file
systems to get the users all happy again).  They never have denied they
will have to replace it again.  It is a standing (not funny) joke we have.

Turn down AFTERBURNER and remove NOVA

We have gone to using the CDC drives as an on-line backup and for material
that does not change often.  In that way, if we have another failure, we do
not loose any work.  My own feelings are that if you need lots of tmp,
swap, or "I don't care if I loose it" type space, the 9771's are just what
you need.  

We have also had bad luck with the CDC "shoebox" FSD about 500Meg drives.
The 2 I have on another system have been replaced 2 times each in the last
year.  In another building here at the labs they have one that gets
replaced every 4-5 months.  At the time of my inital replacement the
service rep. stated that there was an informal trouble report for drives
HDA's under 20,000 and to be forwarned that crashes were coming.  Between
the first and second set of replacments the number moved from 20k to 35k to
50k and to it's present (I think) resting place of 70000.

If someone was to GIVE me a CDC drive I would have to think about it for a
long time as to if I would use it or not.  I do not like to spend time 
recovering from crashed heads.  They would probably have to throw in some
other incentive before I would go for it.

FLAME OFF

These views are mine and not those of the Lab.  With the problems we have
had many people in my group will agree with my comments.  If anyone wants
further comments please feel free to call or email.

Gene Rackow                              ARPA: rackow@anl-mcs.arpa
Mathematics and Computer Science         Voice: 312-972-7126
Argonne National Lab
Argonne Il.  60439