[comp.sys.dec] Decnet Drops Connect

tundra@ux1.cso.uiuc.edu (John Kemp) (08/03/90)

Help!

We just installed a new repeater and are having connections dropouts.
We have a MicroVAXII and Vaxstation 2000 in a Local Area Vax Cluster.
The dropouts appear to occur randomly 4-10 times a day/night.
Basically, the Vaxstation is losing it's connection to the MicroVAX.

The repeater repeats signals from our gateway and MicroVAX
which are on the thick backbone, to twisted pair lines coming out
of the repeater.  The Vaxstation is connected to the repeater
via twisted pair and a TP-THIN transceiver.  The cable lengths
are relatively short so I don't think that is the problem.

Anybody had any luck fixing this kind of problem?  The
repeater is a 12-port Plexcomm AUI-TP repeater if it's
any help.  Operator network messages list below...

Thanks in advance,
--------  john kemp            (  (  )_  internet - kemp@uiatma.atmos.uiuc.edu
  -----                       (  (   __)   decnet - uiatmb::kemp
   ---    univ of illinois   (_ (   __)    bitnet - {uunet,convex}
   --     dept of atmos sci  .(____).               !uiucuxc!uiatma!kemp
   -      105 s gregory ave    ...          phone - (217) 333-6881
    -     urbana, il 61801    ...             fax - (217) 444-4393


********* TYPICAL LIST OF DROPPED CONNECTIONS ****************

2-AUG-1990 12:47:53.74 (from node UIATMC at 2-AUG-1990 12:47:52.26)
12:47:43.56 Node UIATMC (csid 00010002) lost connection to node UIATMB

2-AUG-1990 12:47:53.84 (from node UIATMC at 2-AUG-1990 12:47:52.35)
12:47:52.18 Node UIATMC (csid 00010002) re-established connection to node UIATMB

2-AUG-1990 12:51:53.60  % 
12:51:53.50 Node UIATMB (csid 00010003) lost connection to node UIATMC

2-AUG-1990 12:51:54.05  %
12:51:54.00 Node UIATMB (csid 00010003) re-established connection to node UIATMC

2-AUG-1990 12:51:54.20 (from node UIATMC at 2-AUG-1990 12:51:52.70)
12:51:51.56 Node UIATMC (csid 00010002) lost connection to node UIATMB

2-AUG-1990 12:51:54.33 (from node UIATMC at 2-AUG-1990 12:51:52.78)
12:51:52.64 Node UIATMC (csid 00010002) re-established connection to node UIATMB

tundra@ux1.cso.uiuc.edu (08/04/90)

In case they are of any interest,
here is a summary of the responses I received on the 
LOST CONNECTION problem so far.  Thanks to everyone for the         
suggestions.  You would have to read a lot of manuals
and have a very sharp eye to come up with such good
ideas.  

I have run sys$system:dtsend tests.  I am currently 
trying the RECNEXINTERVAL parameter increase and looking 
at various memory and io parameters.  Only time will tell 
if these changes are successful.

Thanks again,
--------  john kemp            (  (  )_  internet - kemp@uiatma.atmos.uiuc.edu
  -----                       (  (   __)   decnet - uiatmb::kemp
   ---    univ of illinois   (_ (   __)    bitnet - {uunet,convex}
   --     dept of atmos sci  .(____).               !uiucuxc!uiatma!kemp
   -      105 s gregory ave    ...          phone - (217) 333-6881
    -     urbana, il 61801    ...             fax - (217) 444-4393


From: "Carl Karcher, Room 535 WC, 3-5896" <KARCHER@don.waisman.wisc.edu>
Subject: Re: Decnet Drops Connect (?)

The network is flakey. What the messages mean is SCS communications (not decnet)
has been lost and then restored in less than 20 seconds (if sysgen parameter
recnexinterval is at the default). Cluster traffic (ala SCS) is far more
sensitive to network problems that decnet. If the network disturbance lasts
longer than 20 seconds things will crash. I would increase recnexinterval
(say to 1 minute) until your network problems are solved. This is a dynamic
parameter so it can be done online (.e.g using SYSMAN commands PAR SET then PAR
WRITE ACTIVE commands). In a large network its a good idea to increase
recnexinterval permanently. Otherwise one misconnected PC can crash your LAVC.

Carl Karcher                             Internet: KARCHER@WAISMAN.WISC.EDU
University of Wisconsin-Waisman Center   Bitnet:   KARCHER@WISCPSL Madison,
Wisconsin, USA                  	 PSTnet:   (608) 263-5896

Subject: Decnet Drops Connect (?)
Sender: Craig Paul <PAUL@kuhub.cc.ukans.edu>
Reply-To: paul@kuhub.cc.ukans.edu

What does dtsend tell you about line utilization? If you get less than
6% line utilization you've probably got physical line problems. Since
your explanation makes these two nodes sounds as though it is your
entire network I'd expect line utilizations in the 10s of percents, at
least!