7.3-1 to 5.2-1 Connectivity

[Alfonso Urdaneta]

Hello,

At my employer's location we have installed an Alphaserver ES40, and some older V5.2-1 machines. The ES40 is equipped with two network interfaces, and I am tasked with connecting it to my company's TCP/IP network with one interface, and to the V5.2-1 machines via the other.

I have gotten the ES40 on the TCP/IP network, but my problem is configuring it to talk to the V5.2-1 machines.

I have the machines connected via a hub (not the same as the TCP/IP network of course). I have a coax->cat5 adapter on the 5.2-1 machine which I am using to connect it to the shared hub. I believe that I have decnet running on both machines, but I do not know how to confirm this, nor do I know how to confirm that the machines are even physically connected correctly.

If anyone could point me in the right direction, I would greatly appreciate it. I would like to know:

1) how do I verify that DECNET is running and configured on a given machine
2) How do I verify that the versions of decnet on the machines are compatible with one another.
3) how do I verify that DECNET is associated with a given network interface
4) is there a DECNET equivalent to "ping" that I may use to even confirm that these machines are connected properly.

Thank you.

[Steven Schweda · Accepted solution]

VMS V5.2-1 is _old_. (You probably knew this.)

1. SHOW NETWORK should say something.

2. Should not be a problem.

More complex things may depend on the DECnet
versions involved. DECnet Phase IV uses NCP
for management, Phase V (DECnet-Plus) uses
NCL.

DECnet and TCP/IP can coexist nicely on one
network. You may not have needed to use more
than one network interface.

On the old systems, assuming Phase IV, you
could try commands like these:

$ mcr ncp show exec

Node Volatile Summary as of 20-MAR-2006 20:02:55

Executor node = 1.20 (GIMP)

State = on
Identification = DECnet-VAX V5.5-2, VMS V5.5-2

$ mcr ncp loop node gimp
$
(Fill in your own node name instead of "gimp".)

You should also be able to "SET HOST " and get a log-in prompt. Then try
"SET HOST ".

If Phase V is in use, the vestigial NCP
should say something like this:

$ mcr ncp show exec

Node Volatile Summary as of 20-MAR-2006 20:12:27

Executor node = 1.95 (ALP)

State = on
Identification = DECnet-OSI for OpenVMS

[Volker Halle]

Alfonso,

welcome to the OpenVMS ITRC forum.

Let's look at this from the lower layers first:

$ ANAL/SYS
SDA> SHOW LAN

should show your LAN adapters and the various protocols running on them.

SDA> SHOW LAN/COUNTERS

will show the counters for the LAN adapters and their protocols.

With these commands, you could at least see, whether packets are being rcvd/sent. I cannot verify, if this command works on VAX/VMS V5.2-1

On the VAX, also try MC NCP SHOW KNOW LINE COUNT and MC NCP SHOW KNO CIRC COUNTER to view the counter of the DECnet line/circuit.

There are also NCP LOOP LINE and LOOP CIRCUIT commands, but they need some pre-requisite settings, we can leave this for later...

With MC NCP SHO EXE you'll see the DECnet version (i.e. Identification = ...) both on your VAX and your ES40 and once you provide that info, we can suggest the correct commands for your system's DECnet version.

Volker.

[Alfonso Urdaneta]

> VMS V5.2-1 is _old_. (You probably knew this.)

Yeah, I know :) We manufacture test sets and hardware for an automated test station that has a VAX built in to it for the processing - kind of like the test racks the mechanic hooks up to fix your car, but ours repairs big things like radar components.

> 1. SHOW NETWORK should say something.

$ show network

Product: DECNET Node: mynode Address(es): 1.1
Product: TCP/IP Node: mynode.mywork.com Address(es): 192.168.136.60


> DECnet and TCP/IP can coexist nicely on
> one network. You may not have needed to use
> more than one network interface.

That's actually quite interesting, I didn't think that was possible. The machine had two interfaces when it was handed to me, so that's just how I assumed it had to be.

> $ mcr ncp show exec

Node Volatile Summary as of 21-MAR-2006 14:29:24
%NCP-W-SYSMGT, System-specific management function not supported

> $ mcr ncp loop node gimp

$ mcr ncp loop node gimp
%NCP-W-SYSMGT, System-specific management function not supported

> You should also be able to "SET HOST
> " and get a log-in prompt. Then
> try "SET HOST ".

I can do "set host " and get a login prompt, but "set host " or "set host 13.46" both just hang.

[Alfonso Urdaneta]

> welcome to the OpenVMS ITRC forum.

thanks !

> $ ANAL/SYS
> SDA> SHOW LAN

-- LAN Device Summary 21-MAR-2006 14:40:03 --

LAN block address = 814D3800 (2 stations)
LAN block flags = 0000000D REG_VCI,LAN_INIT,CNM_INIT

-- EIA Device Summary 21-MAR-2006 14:40:03 --

LSB address = 827A6000
Device state = 00000003 RUN,INITED

UCB UCB Addr Fmt Value Client State
--- -------- --- ----- ------ -----------
EIA0 814BDF40
EIA2 814F36C0 802E 08-00-2B-80-3C DNAME 0017 STRTN,LEN,UNIQ,STRTD
EIA3 814F3200 Eth 80-3C DNAME 0017 STRTN,LEN,UNIQ,STRTD
EIA4 81597A40 802 FE DECNETV 0017 STRTN,LEN,UNIQ,STRTD
EIA5 81593A40 Eth 60-03 DECNET 0017 STRTN,LEN,UNIQ,STRTD
EIA6 8163BC80 802E 08-00-2B-80-3E DTIME 0017 STRTN,LEN,UNIQ,STRTD
EIA7 81657C00 Eth 08-00 IP 0015 STRTN,UNIQ,STRTD
EIA8 8169EC80 Eth 08-06 ARP 0015 STRTN,UNIQ,STRTD
EIA9 8169F4C0 Eth 86-DD 0015 STRTN,UNIQ,STRTD

-- EIB Device Summary 21-MAR-2006 14:40:17 --

LSB address = 827B0000
Device state = 00000003 RUN,INITED

UCB UCB Addr Fmt Value Client State
--- -------- --- ----- ------ -----------
EIB0 814BB080
EIB2 814E21C0 802E 08-00-2B-80-3C DNAME 0017 STRTN,LEN,UNIQ,STRTD
EIB3 814F2EC0 Eth 80-3C DNAME 0017 STRTN,LEN,UNIQ,STRTD
EIB4 81595A40 802 FE DECNETV 0017 STRTN,LEN,UNIQ,STRTD
EIB5 81591A40 802E 08-00-2B-80-3E DTIME 0017 STRTN,LEN,UNIQ,STRTD
EIB6 816A8440 Eth 08-00 IP 0015 STRTN,UNIQ,STRTD
EIB7 816A8C80 Eth 08-06 ARP 0015 STRTN,UNIQ,STRTD
EIB8 816A94C0 Eth 86-DD 0015 STRTN,UNIQ,STRTD

> SDA> SHOW LAN/COUNTERS

-- EIA Counters Information 21-MAR-2006 14:40:55 --

Octets received 634759472 Octets sent 81219747
PDUs received 4577290 PDUs sent 931696
Mcast octets received 508018850 Mcast octets sent 30031372
Mcast PDUs received 4068327 Mcast PDUs sent 584592
Unrec indiv dest PDUs 0 PDUs sent, deferred 12082
Unrec mcast dest PDUs 119943 PDUs sent, one coll 6493
Data overruns 0 PDUs sent, mul coll 9047
Unavail station buffs 0 Excessive collisions 0
Unavail user buffers 0 Late collisions 0
CRC errors 0 Carrier check failure 0
Alignment errors 0 Last carrier failure None
Rcv data length err 0 Coll detect chk fail 0
Frame size errors 0 Short circuit failure 0
Frames too long 0 Open circuit failure 0
Seconds since zeroed 4652096 Transmits too long 0
Station failures 0 Send data length err 0

-- EIA Counters Information (cont) 21-MAR-2006 14:40:55 --

No work transmits 0 Ring avail transitions 0
Buffer_Addr transmits 0 Ring unavail transitions 0
SVAPTE/BOFF transmits 0 Loopback sent 0
Global page transmits 0 System ID sent 15600
Bad PTE transmits 0 ReqCounters sent 0
Restart pending counter 0 Internal counters size 128
+00 CFG cmd buffer VA 8B176000 +50 Map regs allocated 0
+04 Cmd buffer list VA 8B176020 +54 Rcv buffers mapped 0
+08 Receive ring VA 8B17A020 +58 Xmt segments mapped 0
+0C Stat buffer VA 8B183020 +5C Page boundary xmts 1903
+10 LSB size 7704 +60 PTE to PFN translates 0
+14 Driver version 24 +64 PTE to PFN 2nd page 0
+18 Driver flags 00001814 +68 Chained VCRP xmits 366563
+1C Device ID B0DD0004 +6C Too few segment xmits 0
+20 Soft errors 0 +70 Transmit copies 0
+24 Hardware errors 0 +74 Zero length DCBEs 0
+28 Nway resets 0 +78 Transmit timeouts 0


> On the VAX, also try MC NCP SHOW KNOW LINE
> COUNT and MC NCP SHOW KNO CIRC COUNTER to
> view the counter of the DECnet line/circuit.


$ mc ncp sho exe
Node Volatile Summary as of 21-MAR-2006 14:42:21
%NCP-W-SYSMGT, System-specific management function not supported

> both on your VAX and your ES40 and once you
> provide that info, we can suggest the
> correct commands for your system's DECnet
> version.

That's on the alpha - some dude is running tests on the station, so I can't get to it right now. Can't really do much right now, I'm just trying to figure out the OpenVMS docs... :\

[Alfonso Urdaneta]

Was able to sneak onto the vax

$ mc ncp sho exec

Executor Node = 13.46 (SKP046)

State = On
Identification = DECnet-VAX V5.2-1, VMS 5.2-1
Active links = 6

> MC NCP SHOW KNOW LINE COUNT

prett much everything was zero except for the multicast stuff.

> MC NCP SHOW KNO CIRC COUNTER

Again, everything was zero except for "Circuit down" (56), data blocks sent (11,671) and bytes sent (700,260).

"mcr ncp loop node skp046" appears to do nothing. well, I get a linefeed and back to the prompt anyway...

[Steven Schweda]

> Product: DECNET Node: mynode Address(es): 1.1

That looks like Phase V (DECnet-Plus).

> Executor Node = 13.46 (SKP046)

That looks like Phase IV.

1.1 and 13.46 are in different DECnet areas
(1 v. 13). I don't do this, but I believe
that it requires an area routing node
somewhere to get from one DECnet area to
another.

Who decided what to use for the DECnet
addresses, and why? If you can, you might
change the 1.1 to 13..

> The machine had two interfaces when it was
> handed to me, so that's just how I assumed
> it had to be.

Trust no one, I always say. My main Alpha
has multiple Ethernet interfaces, but
currently only (the built-in) EWA is
connected, and it's doing DECnet and TCP/IP
(and AppleTalk, actually). Having multiple
protocols on one interface is pretty normal.

You might decide to keep the DECnet traffic
segregated from other stuff, but it's not an
actual requirement for normal operation.

> "mcr ncp loop node skp046" appears to do
> nothing. well, I get a linefeed and back
> to the prompt anyway...

I believe that that's success. Try it with
a known-bad name or address to see a failure.

[Alfonso Urdaneta]

So it appears that I can't change network settings because NCP won't do anything, and I can't run that configure file either, because it in turn needs NCP.

Now to figure out what I don't have NCP...

*sigh*

[Steven Schweda]

> [...] I can't run that configure file [...]

Which "configure file"?

On the new Alpha, if DECnet-Plus was
installed (as appears likely), the
configuration procedure is
SYS$MANAGER:NET$CONFIGURE.COM

It uses NCL (which is appropriate for phase
V), not NCP (which would be appropriate for
phase IV). It's relatively straightforward.
I'd vote for the "Perform an entire
configuration" option.

DECnet-Plus includes a vestigial NCP, but it
can't do much. And NCL is entirely
different, so NET$CONFIGURE.COM is usually
easier to deal with.

[Volker Halle]

Alfonso,

to make things easier, wouldn't it be possible to hook your VAX to the same switch/hub your ES40 EIA LAN interface is connected to ?

DECnet end-nodes in different areas (1.x and 13.x) should be able to communicate, if there is no DECnet router on the LAN - they just send out the DECnet packet on the LAN with the correct destination address. To determine, if there is a DECnet router on the LAN try: MC NCL SHOW ROUTING CIRCUIT csmacd-0 ADJ * ALL on your ES40 (if there is no csmacd-0 routing circuit, use MC NCL SHOW ROUTING CIRCUIT * to find the correct name). If there is a router, you would need area routing etc.

The RECEIVE line and circuit counters on your VAX are all ZERO, right ?!

Trying to set up DECnet to be able to communicate via EIB0 to your VAX may be possible, but would require some reconfiguration.

You didn't show the counters for EIB0 on ES40.

Volker.

[Colin Butcher]

Hello Alphonso,

From what I see you're running DECnet Phase IV on the VAXes and DECnet Phase V on the Alpha with Phase IV compatible addressing enabled on adapter EIA.

Can you check which devices the VAXes are running DECnet on? MCR NCP SHO KNOWN LINES and MCR NCP SHOW KNOW CIRCUITS should show the line / circuit name. If the VAXes have multiple adapters then either they will need a routing licence or they should only have DECnet running on one of the adapters. Back then most VAXes came with a single 10Mbit/sec interface.

So, you need to ensure that the VAX ethernet adapters and the Alpha's EIA adapter are on the same physical LAN segment, then it should busrt into life. First try connecting by address, then by name.

You may need to reconfigure which adapter runs in Phase IV compatible mode - use @SYS$MAANGER:NET$CONFIGURE ADVANCED and make changes to the interfaces. Be careful not to enable Phase IV compatible addressing on all adapters unless you know that they go to physically separate (not bridged) LAN segments. Phase IV addressing causes the adapter to run with a MAC address based on the DECnet address of the host - so ALL adapters in a machine can end up running with the same MAC address. Not a problem, provided that you understand it.

You may need to fix the node name to address mapping (NCP SET / DEFINE NODE on Phase IV. MCR DECNET_REGISTER on Phase V - and don't forget the infamous NCL FLUSH SESSION CONTROL NAMING CACHE ENTRY "*" on Phase V afterwards).

If they're all on the same LAN segment then you don't need a router to communicate between DECnet areas. It's good practice to have one - I generally use an old DECbrouter 90 or similar.

You might find this useful for some of the background: http://h71000.www7.hp.com/openvms/journal/v5/decnet.pdf

Cheers, Colin.

[Alfonso Urdaneta]

While playing around with configurations, etc., I't noticed somethign weird. Why do these commands show different values, and could this be part of my problem?

$ sho network

Product: DECNET Node: INDRA0 Address(es): 1.1
Product: TCP/IP Node: indra0.indra.cc Address(es): 192.168.136.60
$ mc ncp sho exec


Node Volatile Summary as of 23-MAR-2006 13:03:50

Executor node = 13.37 (INDRA0)

State = on
Identification = DECnet-OSI for OpenVMS

[Karl Rohwedder]

Whar does

$ MC DECNET_REGISTER SH NODE INDRA

show?

regards Kalle

[Volker Halle]

.. and what's shown by

$ SHOW NET/FULL

$ MC NCL SHOW ROUTING PHASEIV ADDR

Volker.

[Alfonso Urdaneta]

http://www.red82.com/openvms/info.txt

(has the results of the commands that you guys just suggested).

http://www.red82.com/openvms/counters.txt

has the counter information, which I failed to fully provide earlier when asked.

http://www.red82.com/openvms/layout.txt

wooohooo, ascii art !

[Alfonso Urdaneta]

oh, and here is the log of the configuration.

For the life of my I can't figure out what I'm doing wrong.

http://red82.com/openvms/config.txt

[Volker Halle]

Alfonso,

the fact that SHOW NET shows the address as 1.1 seems to be caused by SCSSYSTEMID of your ES40 being set to 1025 (check with MC SYSGEN SHOW SCSSYSTEMID). In general, one would use the DECnet address as the SCS system-id, i.e. 13*1024+37 = 13349.

As your DECnet address is 13.37 on the ES40 and 13.46 in the VAX, you don't need an area router. What you need is a DECnet Phase IV compatible address on EIB0 (routing circuit csmacd-1).

The counters on EIB0 indicate, that there never has been any packet received via this LAN interface during the system uptime of about 40 minutes.

You should first enable DECnet Phase IV protocol on EIB0:

$ mc ncl
NCL> SHOW csma-cd station csmacd-1 comm port

this should show EIB - if true, then:

NCL> disable routing circ csmacd-1
NCL> set routing circuit csmacd-1 enable phaseiv address true
NCL> ENABLE routing circuit csmacd-1

or answer YES to the folloing question in NET$CONFIGURE:

* Enable Phase-IV Addressing on Routing Circuit 'CSMACD-1'? [NO] : YES

You need a DECnet Phase IV compatible address to be able to communicate with your VAX !

Volker.

[Volker Halle]

Alfonso,

looks like your node may have been a router before:

* What type of node (Endnode or Router)? [ROUTER] : endnode

If you have an appropriate DECnet license, you can also configure it as a router again.

But the KEY point is still a Phase IV address on CSMACD-1 !

Volker.

[Alfonso Urdaneta]

> But the KEY point is still a Phase IV
> address on CSMACD-1 !

good eye, for the life of me I can't see how I missed that so many times in the past. redid the configuration again.

http://red82.com/openvms/config2.txt

> looks like your node may have been a router
> before:

From reading the docs, I got the impression that I did not need to configure it as a router, which is why I chose to set it to be an endnode. Your message suggests that I may have a DECNET license that only allows it to be configured as a router. How do I verify this?

> the fact that SHOW NET shows the address as 1.1 seems to be caused by
> SCSSYSTEMID of your ES40 being set to 1025 (check with MC SYSGEN SHOW
> SCSSYSTEMID). In general, one would use the DECnet address as the SCS
> system-id, i.e. 13*1024+37 = 13349.

I'll get on that, thanks.

[Ian Miller.]

Check your DECnet licence. DVNETEXT allows routing, DVNETEND does not.

[Alfonso Urdaneta]

It's a DVNETEND license.

[Jim_McKinney]

As Ian said, DVNETEND permits the system to act only as an end node. As long as licensing is in question, you should also verify that the license has adequate units. On an ES40 I believe that you should find that DVNETEND requires 1050 units ("$ SHOW LICENSE/UNIT" will list each licenses type and I think you'll find that your DVNETEND is a type H). The "$ SHOW LICENSE" command needs to show you that there are at least that many units present.

[Alfonso Urdaneta]

looks good. Time to go bang my head againts the monitor some more...

DVNETEND
Producer: DEC
Units: 1050
Version: 0.0
Release Date: (none)
Termination Date: (none)
Availability: H (Alpha Layered Products)
Activity: 0
MOD_UNITS
ALPHA

[Robert Brooks_1]

Unless you specifically need the OSI transport (perhaps for DECnet over TCP/IP), you can make your life much simpler by installing DECnet Phase IV on your ES40.

While there is nothing particularly wrong with DECnet/OSI (or DECnet-Plus, as it's now called), it take a bit more time to understand its management interface.

As a customer, I began using Phase V with the notorious "Extensions to DECnet-VAX V5.4" in 1992, and I actually used the OSI tranport (for use with X.400), so I do understand quite well how to manage Phase V. Still, for a small network of a few nodes, being familiar with a single management interface
(NCP) will be easier for you to understand than attempting to understand NCL and the various scripts used.

-- Rob (VMS Engineering)

[Alfonso Urdaneta]

I wish I could. Unfortunately we don't have a PhaseIV license, and my vendor has told me that to get one will I'll have to spend ten thousand dollars.

On a postive note, I have now gotten the SCSSYSTEMID set properly, so it now matches the decnet address that I assigned the ES40.

If I can do (on the alpha) "set host 13.37" and get the same result as "set host 0", is it safe to assume that my configuration is correct and it is a hardware issue?