Installing DECNet Plus

[Steve Calnek]

Gentleman, I have an odd issue that I'm hoping you can help with.

I have an upcoming project where I need to install DECNet Plus on to a production system running OpenVMS 8.3 (Alpha). Having read the manuals I decided that I should practice and document the steps needed. To that end I did a fresh install of 8.3 on a DS10 I had available. I did the install and excluded DECNet. Once the install was done and the system was available on the network via TCPIP I shut the system down, booted from the 8.3 install CD and installed DECNet Plus from the install layered products menu. A reboot and autogen later I ran NET$CONFIGURE and DECNet_REGISTER assigning the node name DRSA01 with the Address 1.2.

Everything works! I was patting myself on the back for getting it in one when I noticed that when I enter $SHOW NET the DECNet Address is displaying as 63,1022.

I've retraced my steps and find NET$CONFIGURE is correct and DECNet_REGISTER shows the assigned 1.2.

Any idea's on why the displays wrong and how I might correct it.

Thanks

Steve/

[Hoff · Accepted solution]

Confirm the system parameters SCSSYSTEMID and SCSNODE are set correctly. The former is set to (node_area * 1024 + node_number, and the latter is the node name.

Here's a list of where the host name can lurk:

http://labs.hoffmanlabs.com/node/589

[Volker Halle]

Steve,

did you do another reboot after NET$CONFIGURE ?

If not, did you manually run @SYS$STARTUP:SYS$NET_SERVICES.COM ? This procedure will register information about all network stack (TCPIP and DECnet) for SHOW NETWORK to display.

Volker.

[Colin Butcher]

That sounds as if the SCSSYSTEMID value doesn't match the DECnet Phase IV compatible address. For Phase IV the DECnet address and the SCSSYSTEMID values should match, using the calculation SCSSYSTEMID = (AREA * 1024 + NUMBER).

All this goes back to when clustering and DECnet were both integral parts of the OS. With Phase V and the removal of the requirement that clustering required DECnet it's not so tightly coupled, but there are things that expect the SCSSYSTEMID field and the DECnet Phase IV address to match.

It's easy enough to fix - modify MODPARAMS.DAT and set the SCSSYSTEMID value in there, then use AUTOGEN and reboot the machine.

The good thing about Phase V is that you can control which NICs have Phase IV compatible addresses (and thus get a Phase IV compatible MAC address change). You can also use "DECnet over IP", where the TCP/IP stack gets used as a pseudo transport for DECnet traffic. If you use "DECnet over IP" only, without running NSP or OSI over a physical NIC, then you don't need a DECnet Phase IV compatible address at all.

This article from the OpenVMS Technical Journal V5 might help with some of the background if you're interested: http://www.downloads.xdelta.co.uk/vmstjv5%20feb2005/decnet%20article%20vms%20tj%20v5%20feb2005.pdf

Cheers, Colin (http://www.xdelta.co.uk).

[Steve Reece_3]

As others have suggested, try rebooting to make sure that the data fields are populated correctly for which networks are running.

After this, you can also try doing SET HOST 1.2 and see that you get to your node with no errors or warnings of host name not being translatable (I don't remember the exact phrase that the error message uses). If you can communicated with the node 1.2 then you're ok. If not then try:
$ mc ncl flush session control naming cache entry "*"

to make sure that there's nothing odd that's in cache. TCP/IP drops all of the host information and reloads it when it restarts. DECnet Plus tends to keep stuff in cache instead.

Steve

[Steve Calnek]

Hi Gents,

Thanks for the help. Hoff's document gave me my answer so top points to him. Colin your answer was on the money but second to get to me so I gave you points to.

Solution was,
Corrected SCSSYSTEMID based on the calculation provided followed by an AUTOGEN.

Thanks Again

Steve/