DECnet OpenVMS and Cisco

[geir_2]

Hi,

Thanks for the answer. I have done what you asked me to do (attached document).

I'm obviously not a DECnet expert, the DECnet host is running DECnet IV, maybe the router is confused about it, and should be configured as Level IV area router

Geir

[Volker Halle]

Geier,

the EWA0 LAN counters on FBUFS1 look o.k. - this interface does successfully receive and send packets.

The same interface counters shown for EWA0 on node NITFS1 earlier did show NO packets sent and all LAN transmits failing with carrier check failed. As both nodes are running the SAME version of OpenVMS, there must be something wrong with the HARDWARE (LAN interface, network cable, switch port). You may argue, that IP works on NITFS1, but DECnet doesn't. The router seems to see DECnet HELLO messages from 2.81 - I also assume these messages stop, when you disconnect the network cable to NITFS1.

Please post SDA> SHOW LAN/COUNT from NITFS1 for all protocols. If the interface tells, that it failed to send any packets, IP packets should get the same error. The interface may report a 'carrier check failure' on every transmission, while it actually sent the data out on the wire. The upper layers may react differently on this status.

Volker.

[Ian Miller.]

You have spent some time (and therefore money) on this problem in this forum.

If these systems are commercially significant it would be more cost affective for you to buy few days consultancy from someone who knows about this sort of thing who can fix the problem and teach you how to deal with it in future.

There are several people who participate here who could do this (some of who have been answering your questions - I'm not one of those people).

[geir_2]

Volker,

The output to the command is attached below. Hope the information is to help.

Regards geir


PS:
The OpenVMS host is running Phase IV, and the router is identical at both sites with the following commands:


decnet routing 2.1
decnet node-type area

interface gig 0/0
decnet cost 1

interface tunnel 12
decnet cost 10

[geir_2]

Hi again,

I forgot to post the output


Geir

[Volker Halle]

Geir,

you seem to have posted the FBUFS1 counters again. I was interested to see all LAN coutners from NITFS1...

I still believe the following to be true:

There is something wrong with the network hardware on NITFS1. This node does NOT successfully send any network packets.

Volker.

[geir_2]

Hi,

Sorry, this time I will post the information you asked for. It's correct, what you said about the interface. It seems to be something wrong with it.

But, how is it possible for IP to work 100%?

I also though it could be something wrong with speed and duplex. The switch is Cisco 3560 (C3560G-IP-BASE-M, VERSION 12.2)

The router is cisco 28xx with IOS 12.4.

Is it possible that DECnet use different settings for speed and duplex than IP??
Is it possible to use a show command that confirm it?


Regards Geir


PS:
We have a software agreement (actual server)with HP on this server, but first of all I want to verify all possibilities

[Volker Halle]

Geir,

none of the protocols running on EWA have successfully send ANY packet - according to the counters. I'm assuming that the packet sent counters will only be incremented after a successful transmission of the packet. If the LAN interface reports 'carrier check failure' as the packet transmission status, the LAN driver may not increment the 'sent' counters ! The packets still seem to be sent on the wire...

Try using a working network cable first, which you have verified to work (with the other OpenVMS system ?).

Volker.

[Ian Miller.]

If you have paid for support for HP then use it. They can spend more time and visit your site if needed to fix the problem.

It will be faster and cheaper for you.

[geir_2]

Hi,

It's obviously not a HW problem. Early this morning I did some investigation. I did following:

If I did following command ana/sys I discovered some mismatch between FBUFS1 and NITFS1. On FBUFS1 DECnet was listed first, On NITFS1 IP was listed first. Someone told for a long time ago that DECnet must be configured/started first. Therfore I executed following commands

a) mc ncp set exec state shut
b) @ucx$config and stopped IP
c) @netconfig
d) @ucx$config and started IP


And now, suprisingly both DECnet and IP was
working 100%. If I reboot NITFS1, the same procedure happen. So I wondered what I must do to survive a reboot on NITFS1.

Regards Geir

[Robert Gezelter]

Geir,

It sounds unusual that the startup of the networks in the incorrect order would cause these symptoms, much less that it would ever have worked. However, working in the forum (as opposed to actually seeing -- locally or remotely -- the machine), I must admit that things can appear to present in strange ways.

The networks are started from within SYS$MANAGER:SYSTARTUP_VMS.COM. The lines to invoke NETCONFIG and the IP startup procedure may also be in subsidiary files (this is the default way of doing it, the files can also be directly invoked by the STARTUP procedure using entries in the STARTUP database, a procedure that I have used on may sites to speedup restart, and have spoken about at several symposia).

Find the references and verify that they are in the desired order. DECnet needs to be started first. The reason for this is that DECnet changes the MAC address of the ethernet adapter used to an address in the DECnet reserved range The precise MAC address is an amalgam of the DECnet defined (and officially registered prefix) and the DECnet node number).

I hope that the above is helpful.

- Bob Gezelter, http://www.rlgsc.com