Conference netcad::hub_mgnt

3020.0. "Concentrator 900's resetting" by WARFUT::FEENEY () Thu Nov 30 1995 10:47

I wonder if someone could help with a problem of DECconcentrator 900MX
intermittently resetting themselves?

The rev levels are HW=V2, RO=V1.2, SW=V3.0.0

The failing concentrators errorlogs all showed an entry of
       "Stop thrash: cleared Nonvolatile data"
and several entries of
       "Trap@322 in sys-ip-stack-cfg.c"

Six of the eight concentrators were each reseting at the same time, so it
appears to be driven by a network "event". The two which didnt have the problem
had IP addresses so we tried setting the rest with IP addresses and they have
stayed up for a week OK now. The customer doesnt consider this a fix and would
like to know whats happening, I wouldnt mind knowing myself either.

The customer has recently installed the FDDI network as in the following diagram:

           4 rings                      4 rings
         _______________________Giga________________________
	|                      switch                      |
        |                                                  |
     2 Hub900's                                        1 Hub900
     8 Switch900EF                                     4 Concentrator900MX 
  2 Switches on each ring                              1 on each ring
        |                                                  |
        |                                                  |
        |                                              2 Hub900's
        |                                              8 Switch900EF
        |                                           2 Switches on each ring
     1 Hub900                                              |
     4 Concentrator900MX                                   |
     1 on each ring                                    2 Hub900's
        |                                              8 Switch900EF
        |                                           2 Switches on each ring
        |                                                  |
        |                                                  |
        |_______________________Giga_______________________|     
                               switch

Has anyone seen anything similar or could someone tell me the significance of
the Trap errlog message please?

		Thanks in advance

				-Nick

Hi Nick.  There was a problem at one point where the concentrator
(with no IP address) would run out of transmit buffers and reset.
If I remember correctly, this was found to be caused by a particular
type of UDP packet.  Version v2.8 and higher had the fix for this
bug.

The concentrator is only going to look at an IP packet if the destination
address is the concentrator's IP address or if the destination
address is a broadcast address.  Since the concentrator's without
IP addresses are the ones crashing, it must be some sort of broadcast
packet.  It would help me a lot if you could tell me the broadcast
traffic on the wire at the time the concentrators reset.  Can you
do this?

In the mean time, I can make available to you a new concentrator image that
we have in system test right now.  You can try this new image on a couple of
the concentrators and see if they have the same problem.  If they do, I will
stop system test until we get it fixed.

Dennis