2 DEGPA on ES40 switching DECnet

[Jeff Lanka]

We have 2 DEGPA on ES40's running DECnet phase 4 VMS 7.3-1 and I want to be able to switch from one to the other without rebooting. I use the NCP (Phase 4) utility and simply set the state off on line/circuit on EWA-0 and set the state on line/circuit on EWA-1.

On most of the nodes I'll get an IVADDR error when I try to bring up the other line. The only thing that will allow me to switch on them is a reboot.

But on 2 of the nodes, when the machine boots the inactive DEGPA is in "disabled by software" state whereby it has a blinking green light. These machines ARE able to switch on the fly successfully.

Question: How do we get all our machines to come up with the inactive DEGPA in the "disabled" state?

[Antoniov.]

Hi Jeff,
I think it's better keep all lines to state on and switch only the circuit.
Example:
NCP>SET KNOWN LINE STATE ON
NCP>SET CIRC EWA-0 STATE OFF
NCP>SET CIRC EWA-1 STATE ON
Above command change only volatile database, obviously; so after switch you are able to make change on permanent database:
NCP>DEF KNOWN LINE STATE ON
NCP>DEF CIRC EWA-0 STATE OFF
NCP>DEF CIRC EWA-1 STATE ON

If previous command don't work, you could make change on permanent database, shutdown decnet e restart it:
NCP>SET KN CIRC STATE OFF
NCP>SET KN LINE STATE OFF
NCP>DEF KNOWN LINE STATE ON
NCP>DEF CIRC EWA-0 STATE OFF
NCP>DEF CIRC EWA-1 STATE ON
NCP>SET EXEC STATE SHUT
$ @SYS$STARTUP:STARTNET

H.T.H.
Antonio Vigliotti

[Jeff Lanka]

I had tried that, and when I booted the machines which have the problem of the IVADDR when I switch, they gave me the same IVADDR on boot.

i.e. I defined lines EWA-0 and EWA-1 state on, and defined circuit EWA-0 state on and defined circ EWA-1 state off. Reboot. When DECnet starts on boot, get IVADDR. THen go into NCP and show know lines, just see EWA-0.

[Antoniov.]

Jeff,
I guess you have some h/w trouble or you need configure decnet.
Do you executed @SYS$MANAGER:NETCONFIG.COM?

Here in Italy is evening and I'm going to home. I'll read your answer tomorrow morning; sorry for this black-out.

Antonio Vigliotti

[Volker Halle]

Jeff,

you'll get the IVADDR error, if the DAT (Duplicate Address Test) fails, i.e. there is another interface/node on the LAN using the SAME address (in you case AA-00-04-00-nn-nn). See HELP/MESS IVADDR

Check with SDA, which Physical Address both of your DEGPAs are using: SDA> SHOW LAN/FULL/DEV=EWA and /DEV=EWB

The duplicate address test is run as soon as you NCP> SET LINE EWA-n STATE ON, so you'll get this message, if both LINEs are DEFINED with STATE ON in the permanent database during boot (STARTNET).

Depening on other protocols running on the line to be switched off, the address cannot be changed SDA> SHOW LAN shows you the protocols running on each LAN interface). If you would disconnect the line from the network, you should be able to activate the other line.

Volker.

[Jeff Lanka]

Thats all well and good, and I'll try all the tests as soon as get into work.

But nobody on any forum has been able to answer the question I posed: How is it that 2 of the machines have the inactive DEGPA come up with the blinking green light which the book says means "disabled by software"? Nobody even mentions that. Anyone want to take a shot at that one?

[H_Bachner]

With DECnet Phase IV the MAC address (the "hardware" address seen on the wire) of the NIC is changed to some AA-00-04-xx-xx-xx address, where the DECnet address is encoded in the xx-xx-xx part.

If you switch off the circuit for this NIC, the MAC address is *not* reset (and cannot be reset in some cases, as Volker (hi!) pointed out). Usually, the MAC address is reset when the NIC is power cycled, though you could certainly write a small program which resets the MAC address.

When you enable the second circuit, DECnet tries to set the MAC address of the second NIC to the same AA-00-04-xx-xx-xx address, because the DECnet address is still the same. If both NICs are connected to the same LAN, the result is the IVADDR error message.

What are you trying to achieve? If you tell us a bit of the background, some participant in this forum may be able to suggest an alternative.

Hans

[Volker Halle]

Jeff,

you may want to compare the DEGPA internal counters and status information on the 2 systems behaving differently:

SDA> LAN INTERNAL_COUNTERS/DEV=EWAx

I'm not aware of any method to 'disable' the DEGPA by software.

Volker.

[Jeff Lanka]

I'm aware of the DECnet MAC address mangling. The 2 DEGPA's are on seperate LAN's so that isn't a factor.

AFA the "disabled by software" that is listed on page 1-6 of the "Compaq PCI to Gigabit Ethernet Adapter" PN EK-DEGPA-IN.B01 manual.

We're trying to create a com file which will allow operators to switch from on NIC to the other in case of failure at the switch. We can't take advantage of automatic failover because we're only at 7.3-1

I'm going to try the SDA commands now and will report backt he findings. THanks for all your help so far.

[Volker Halle]

Jeff,

maybe the 2 'separate' LANs are really bridged/switched. In that case, your previous entry (EWA-0 and EWA-1 LINE defined to STATE ON) gives the expected behaviour during boot: SET LINE EWA-1 will fail with IVADDR and you'll only see EWA-0 with NCP SHOW KNO LINE afterwards.

It should be possible to shutdown the DEGPA with a program using the QIO interface ( IO$SETMODE ! IO$M_SHUTDOWN ), but that will stop all protocols. And I don't think there is any utility available in OpenVMS to issue such a command.

Volker.

[Jeff Lanka]

I've got a call in to the network admin to double check that the lans are totally seperate. I'll report my findings back here. Thanks for the help

[Jeff Lanka]

OK, I've got egg on my face. The network admin said the 2 LANS do see each other. Sorry about the misleading info.

New question:

If I do a set circ EWA-0 state off and set line EWA-0 state off, are there still ARP packets flying around that the wire sees? Is there some sort of time delay before I can bring the other NIC up and not get MAC address conflict?

THanks again for all you help.

[Volker Halle]

Jeff,

DECnet does NOT use ARP. It sets the DEGPAs MAC address to AA-00-04-00-nn-nn when you start the LINE (if there is no IVADDR conflict) and it will stay that way (check with SDA> SHOW LAN/FULL/DEV=EWc) - did you read $ HELP/MESS IVADDR ?

There is no utility to 'switch back' the MAC address to the device hardware address. It may be possible to write some code, but you'll have to think about all the other protocols running on that DEGPA.

This kind of scenario is what LAN Failover (new in V7.3-2) is for.

Volker.

[Jeff Lanka]

VolkeOK - I now about the DECnet MAC addrss mangling, and that it will stay that way forever.

My question is:

When I do a

MCR NCP SET CIRC EWA-0 STATE OFF
MCR NCP SET LINE EWA-0 STATE OFF

How long until I can bring another NIC up on that same segment and not get an IVADDR error?

[Jeff Lanka]

BTW, I forgot to say thanks for your replies. I really appreciate all the help I've gotten in this forum.

[Antoniov.]

Hi Jeff,
I guess you try as here:
NCP>SET CIRC EWA-0 STATE OFF
NCP>DEF CIRC EWA-0 STATE OFF
NCP>DEF CIRC EWA-1 STATE ON
NCP>SET EXEC STATE SHUT
$ @SYS$STARTUP:STARTNET

In this way you reboot decnet and you can avoid any time-out trouble.

Antonio Vigliotti

[Volker Halle]

Please note that DAT (Duplicate Address Test) is run when you turn on the LINE, not the CIRCUIT !

But even then, you may NOT be able to achieve what you want. The first DEGPA (EWA-0), which had been running DECnet, will NOT change it's address, it will remain at AA-00-04-00-nn-nn until the next reboot (if you have done a DEFINE LINE EWA-0 STATE OFF). Even restarting DECnet won't help. There is no mechanism to have the interface switch back to it's hardware address, other protocols running on that device may also prohibit an address change !

In your scenario, if you want your operators to turn on another EWA line and circuit, because 'there is a problem with the switch', why not disconnect the network cable completely, this would prevent the IVADDR failure.

Volker.

[Jeff Lanka]

OK, let's just say we're only talking about DECnet, no other protocols are running. So if I set the line and circuit state both to off, it's still communicating on layer 2? i.e. the MAC address is still being somehow broadcast such that the other NIC see's it and gives and IVADDR?

BTW, nobody still has explained to me what that blinking green light "disabled by software" means, and how to achieve it. Thats important because 2 of the nodes in question have that, and they are both able to perform the NIC switch successfully.

Thanks again for all your help.

[Volker Halle]

Jeff,

to find out which Physical Address (MAC address) the DEGPAs in your systems are currently using, use SDA> SHOW LAN/FULL/DEV=EWA (or EWB).

I assume that DAT sends an equivalent of a NCP LOOP CIRC xxx PHYS ADDR AA-00-04-00-nn-nn (nn-nn being calculated from the DECnet Phase IV address area.node). If a loopback response is received, some LAN interface is using this address and IVADDR is returned. HELP/MESS IVADDR explains, that the LAN adapter may continue to respond to the physical address even after the node has been HALTed.

Regarding the 'disabled by software' state, I can only guess, that this state may be reported if NO protocol is started on that DEGPA at all, again check with SDA> SHOW LAN. Still I can't see, how you could be able to successfully start the LINE on the secondary DEGPA, if the first DEGPA is still connected to the same LAN segment and using the DECnet-Phase IV MAC address (please check with SDA).

Volker.

[Jan van den Ende]

Just a wild guess from reading the previous entries and applying some logic reasoning(without any real knowledge and no way to verify):
IF two of your systems are running DECNETplus (aka V5), WITHOUT a V4 alias, will that mean you do not get the translation to an AA- adress?

Any reaction from anyone more knowledgable in this stuff??

Jan

[Jeff Lanka]

The 2 systems that have the NIC's that come up in the "disabled by software" state are both running phase 4. Those are the ones that are able to switch successfully.

[Volker Halle]

Jeff,

what are the physical addresses used by the DEGPAs on those 2 systems where the DEGPAs come up with 'disabled by software' - just check whether the addresses are the actual HW-address or the DECnet-Phase IV address ?

Still puzzled...

Volker.

[Jeff Lanka]

OK, I compared 2 nodes.

Node A: (DOES have blinking green light/disabled by software)

EWA: HW: 00-60-CF-21-19-4A
Physical: AA-00-04-00-CB-DE

EWB: HW: 00-60-CF-21-18-D2
Physical: AA-00-04-00-CB-DE

EWA-0 line & circuit is ON
EWA-1 line & circuit is OFF

Node B: (does NOT have blinking green light/disabled by software)

EWA: HW: 00-60-CF-21-19-3F
Physical: AA-00-04-00-C5-DE

EWB: HW: 00-60-CF-21-04-8D
Physical: 00-60-CF-21-04-8D

EWA-0 line & circuit is ON
EWA-1 line & circuit is OFF

Anything else from that SDA that would help?

Thanks again for all your help.

[Volker Halle]

Jeff,

on node A, both DEPGAs have the SAME physical address, so unless they are connected to separate (non-bridged) LAN segments, this would be 'illegal'. Node B looks 'normal' to me.

I assume EWB0 on node A to be the one with the 'blinking green light'. Maybe it's disabled (by whom ?) becasue of the address conflict. What protocols are running on EWB0 (SDA> SHOW LAN) ? How did EWB0 get the DECnet address ?

Volker.

PS: would it make sense to continue this via mail and then post the results here - once found ?

[Jeff Lanka]

sure my e-mail addy is jefflanka@yahoo.com

BTW, here's the protocol page from SHO LAN/DEV=EWB on Node A:

-- EWB Unit Summary 2-AUG-2004 10:19:05 --

UCB UCB Addr Fmt Value Client State
--- -------- --- ----- ------ -----------
EWB0 814CC680
EWB2 81540D40 Eth 60-03 DECNET 0006 LEN,UNIQ