[comp.os.vms] DEBNT "tricks"

SBAILEY@DOMINOES.UUCP (Scott Bailey) (06/01/87)

Our site has just installed a VAX 8700 (with another just beginning
installation) and, in the time the system has been running, we have had
continuing problems with the DEBNT ethernet interface.  Has anyone been
down this road before?

The VAX is connected to a DELNI (which it shares with a DECserver 100 at
present) and the DELNI is connected to the ethernet backbone.  We run
DECnet, LAT, XNS, PUP and who knows what else on this segment...

We boot the VAX and everything is happy.  Yeah, we get a nasty message from
ETDRIVER saying that it and our DEBNT microcode are incompatible -- DEC
says to ignore it.  Everything runs fine for 10, 12 or so hours and then...

                                hic!

...the DEBNT 'dies.'  LAT drops dead.  Circuit BNT-0 goes into
'synchronizing' state.  The error count on device ETA0 increases at a rate
of approximately 1 error/second.  Worse, BUFIO-type IRPs are being
allocated (and not returned) at the rate of 10-20 packets/minute.
Eventually, the IRP lookaside list fills, expands, fills... expands to
IRPCOUNTV, fills... IRPs start coming out of nonpaged pool, which
fragments, fills, expands, fills...  fatal bugcheck...  The system reboots
and everything is again fine.

We've been talking to both CSC and field service.  They checked that
ETDRIVER was at revision 15, and the DEBNT was at revision 40.  Our 8700
console software was upgraded also (I don't know what rev).  CSC insists
that this is a hardware problem; replacing the DEBNT hasn't changed
matters.  Field Service is now telling us that the console upgrade was only
a part of a fix which includes VMS V4.6 (which, naturally, we haven't got.)

In the meantime, our shiny new system is condemned to a life of frequent
reboots.  Even if we could live without the DEBNT working all the time
(never mind that all of our users are DECserver-based), watching 8-10 Mb of
physical memory getting blown on bogus packets really impacts the
performance of our hog jobs.

This isn't the first VAX 8700 in the world.  Has anybody else run into this?

Thanks for any information or comments,

  Scott Bailey                    SBailey@ESGSDWCO.Xerox.com
  VAX Systems Manager             (Don't believe the return path...)
  Xerox Corp. RE/GSD/WCO          213 / 333-5441