chronic DEC hardware problems

[utzoo!decvax!ucbvax!unix-wizards]

>From wales at UCLA-Security Thu Nov 12 23:58:27 1981
We had a chronic hardware problem here at UCLA during March and April of this
year.  Our 11/780 (we ran UNIX/32V till January 1981, then switched to 4.0BSD)
was getting machine checks and hard UBA errors with a frequency of one every
day or two.  During a six-week period, DEC replaced every board in the machine
with the exception of the MBA boards (some individual boards were replaced two
or three times!); they replaced the UBA and CPU backplanes; and finally, in
response to our total frustration and stated readiness to pursue other vendors
for our future computing needs, DEC replaced the entire CPU with a brand-new
one straight off the assembly line.  Since then, we've had hardly a problem at
all in comparison.

Our configuration was (and still is) an RP06 and a CDC 9766 (the latter being
a 300-Mb drive accessed via a System Industries SI9400 controller via the UBA).
We had more than our share of finger-pointing at the SI-supplied disk (we all
know, of course, that DEC and SI are anything but bosom buddies), and DEC even
spent two or three weeks sending out experts to check our AC power and ground-
ing (hoping to find glitches -- they didn't).  As for blaming the UNIX kernel,
they didn't really do that too much, but we heard quite a lot of "Well, if you
were running VMS, we could look at the error log info, but since you're running
UNIX, we're kind-of groping in the dark".

Our original CPU was an early edition (we got it in the late summer of 1979),
and we had been having chronic UBA problems with it almost since Day One.  As I
stated above, though, we have had essentially no CPU- or I/O-adapter-related
problems since they gave us a new CPU.

-- Rich Wales (wales at UCLA-Security)