Possible duplex mode mismatch condition detected

[Paul Whapshott]

I am receving the following message on an ES45 runing VMS 7.3-2.

This is a newly configured server that I am trying to get onto our network. However I cannot get TCPIP up and running on it. I believe because of this error.
The port speed has been set to 100 FULL DP on the ES45 and switch.

Any ideas ?


Regards,

Paul

[labadie_1]

What gives

$ mc lancp sh dev/param

[Bojan Nemec]

Paul,

Check once again that both sides are set at same speed.
Auto negotiation must be also the same on both sides (enabled on both sides or disabled on both sides).

You can experiment by setting the speed and auto in LANCP ($ MC LANCP)

Bojan

[Richard Brodie_1]

I'd start by loooking at the outputs of:

MC LANCP SHOW DEVICE/PAR,
MC LANCP SHOW DEVICE/COUNT, and
TCPIP SHOW INTERFACES.

[Paul Whapshott]

Just resolved the issue.

The ES45 came with 10 half duplex. I changed to FASTFD at >>> then did an init.

I have now just switched the box off and back on and now all is working.

Question is Why did I have to switch off/on to get this working ?


Paul

[Robert Brooks_1]

Standard refrain:

The VMS Engineer who writes and maintains the ethernet drivers has repeatedly stated that with a modern switch and recent versions of VMS (V7.3-2 with patches is considered "recent" in this context), one should always set both ends (the switch and the adapter) to autonegotiate for best performance.


-- Rob

[John Gillings]

Paul,

"working" and "correct" are not necessarily the same thing! Duplex mismatches are notorious for apparently "working" even though the settings are incorrect. Failures typically occur when the network is under load.

You should not have needed to power cycle to have the adapter settings changed in the console.

Since most switches default to autonegotiate, it's really much better if all OpenVMS systems are set to auto too.

To see the "discussion" (if any) between your adapter and the switch, use:

$ MCR LANCP SHOW DEV adapter/INTERNAL

(requires PHY_IO).

[Stephen Eickhoff_1 · Accepted solution]

The proper way to change speed to 100 Mbit and full duplex on a running system (during scheduled maintenance interval, of course) is:

MCR LANCP SET DEV xyzn: /full_duplex /speed=100

Once proper operation is confirmed, this can be set in the permanent config:

MCR LANCP DEFINE DEV xyzn: /update /volatile

The HP field engineer will, looking at the issue from a "hardware guy" perspective, usually recommend setting this from the console instead. VMS will use the console settings if nothing is defined in the LANCP database.

I concur with the other responses, in that older versions of VMS on older firmware usually worked best with both the switch port and the NIC locked to the same settings, rather than auto negotiating. The DE500 was particularly notorious for this. We used to have the most trouble with managed switches for some reason. Of course, that's fortunate as in an unmanaged switch you have either auto negotiate, half duplex, or nothing.