topic Re: %EWA0 possible duplex mode mismatch condition detected in Operating System - OpenVMS

%EWA0 possible duplex mode mismatch condition detected

Tim Schofield_1 — Wed, 14 Mar 2007 08:17:35 GMT

Hi,
I just loaded update 900 and a few other patches on some of my Alpha's running openVMS 7.3-2 and we are getting this message occasionally on the consoles (%EWA0 possible duplex mode mismatch condition detected). There does not seem to be any performance issue or strange counter numbers on the Nic cards in question. Is this possibly a condition that was always there and is just being logged now with the new patches. The Alpha's in question are on different networks and in differnet data centers. I looked around for this error but have yet to find anything specific.
Thanks
Tim

Re: %EWA0 possible duplex mode mismatch condition detected

EdgarZamora_1 — Wed, 14 Mar 2007 08:37:21 GMT

More details are needed. Are you getting the message from just one alpha, multiple alphas? Are your NICs gigabit? Output from LANCP and SCACP (if clustered) would be helpful, too. Most likely your problem is a mismatch between NIC and Switch port settings. Even though you don't notice a performance problem now, you most likely have a performance problem under load.

Re: %EWA0 possible duplex mode mismatch condition detected

Ian Miller. — Wed, 14 Mar 2007 08:44:17 GMT

" Is this possibly a condition that was always there and is just being logged now with the new patches. " - quite likely. Recent patch kits have back ported some enhancements in this area from VMS V8.x

Re: %EWA0 possible duplex mode mismatch condition detected

Richard Brodie_1 — Wed, 14 Mar 2007 08:44:57 GMT

It think it's something from 8.2 that's got backported. The doc says it just polls the counters and looks for something that looks like the footprint of a duplex mismatch.

I'd be inclined to look at the output of MC LANCP SHOW DEV/COUNT. If you've looked already and the counters look clean, that's odd.

Re: %EWA0 possible duplex mode mismatch condition detected

Hoff — Wed, 14 Mar 2007 08:55:08 GMT

The NIC duplex settings appear to the driver to be inconsistent; that one end is set half-duplex, and the other is operating full duplex.

This message and this detection is a V8.2 feature, and it looks like it's been backported.

I'd check the settings on the switch and on the card at a start, and verify that they match. LANCP can show the NIC settings.

Certain of the older cards don't negotiate all that well -- the FAQ has the matrix of DE500, and the DE500 series uses EW driver. Some of the cards auto-detect speed, and some auto-negotiate duplex, and some DE500 NICs manage both. And then there can be manual settings on the switch or the NIC which can override this, which can be at the switch, or most any NIC around.

The DE500 rated its own section in the OpenVMS FAQ because of auto-detect and auto-negotiate.

You can use LANCP to set and show the NIC settings, and you can use SRM console environment variables to select the NIC behavior.

Check the NIC and the switch.

Stephen Hoffman
HoffmanLabs

Re: %EWA0 possible duplex mode mismatch condition detected

Tim Schofield_1 — Wed, 14 Mar 2007 09:18:27 GMT

Thanks for the info. It looks like there is a problem with this nic. I did speak with the network people at this site and they said they have the switch port set to 100MB full duplex, no autosense, no autonegotiate.
I cannot get into the alpha at my other site at this time.

AUG$ mc lancp
LANCP> show dev/param ewa0

Device Characteristics EWA0:
Value Characteristic
----- --------------
1500 Device buffer size
Normal Controller mode
External Internal loopback mode
00-16-35-3E-0B-FB Hardware LAN address
Multicast address list
CSMA/CD Communication medium
FF-FF-FF-FF-FF-FF Current LAN address
128 Minimum receive buffers
256 Maximum receive buffers
Yes Full duplex enable
Yes Full duplex operational
TwistedPair Line media type
100 Line speed (mbps)
Disabled/No Failset Logical LAN state
0 Failover priority

Device Counters EWA0:
Value Counter
----- -------
742319 Seconds since last zeroed
632755804 Bytes received
1026765093 Bytes sent
6668058 Packets received
4533370 Packets sent
163935213 Multicast bytes received
4308653 Multicast bytes sent
2429933 Multicast packets received
75139 Multicast packets sent
2 Unrecognized unicast destination packets
371463 Unrecognized multicast destination packets
0 Unavailable station buffers
0 Unavailable user buffers
245 Alignment errors
11122 Frame check errors (14-MAR-2007 10:00:16.22)
0 Frame size errors
0 Frame status errors
0 Frame length errors
0 Frame too long errors
0 Data overruns
0 Send data length errors
0 Receive data length errors
0 Transmit underrun errors
0 Transmit failures
0 Carrier check failures
0 Station failures
0 Initially deferred packets sent
0 Single collision packets sent
0 Multiple collision packets sent
0 Excessive collisions
0 Late collisions
0 Collision detect check failures
1 Link up transitions ( 5-MAR-2007 19:42:17.52)
0 Link down transitions
None Time of last generic transmit error
None Time of last generic receive error
LANCP> exit
AUG$ sho sys/noproc
OpenVMS V7.3-2 on node AUG 14-MAR-2007 10:05:30.43 Uptime 8 14:23:23

Re: %EWA0 possible duplex mode mismatch condition detected

Tim Schofield_1 — Wed, 14 Mar 2007 09:46:09 GMT

Is there a way to tell from VMS if the Nic card is set to Auto or 100mbs full duplex etc at the SRM prompt?
Lancp> sho dev/params shows that it is 100 full duplex and does not say anything about Auto so I am assuming ega0_mode is hard set to 100mbs full duplex.
Thanks

Re: %EWA0 possible duplex mode mismatch condition detected

Steven Schweda — Wed, 14 Mar 2007 12:06:52 GMT

You may simply have some duplex-unrelated
network flakiness which is being
misinterpreted. Around here:

14914392 Seconds since last zeroed
26756338092 Bytes received
36475621960 Bytes sent
107439829 Packets received
163219332 Packets sent
846741906 Multicast bytes received
852344449 Multicast bytes sent
15530177 Multicast packets received
6960234 Multicast packets sent
[...]
0 Alignment errors
0 Frame check errors
[...]

Re: %EWA0 possible duplex mode mismatch condition detected

Volker Halle — Wed, 14 Mar 2007 12:08:45 GMT

Tim,

the LAN drivers keep the most recent messages issued on OPA0: in memory for you to look at with:

$ MC LANCP SHOW DEV/INTERNAL ewa

This information should indicate, whether the NIC was set to auto-negotiate at console level. It also shows the auto-negotiation settings at the NIC level.

The NIC is seeing FC and alignment errors, so there seems to be something wrong...

Volker.

Re: %EWA0 possible duplex mode mismatch condition detected

Richard Brodie_1 — Wed, 14 Mar 2007 12:30:40 GMT

It's much harder to diagnose duplex mismatches reliably from the full duplex end. With just CRC/Align errors it could just be general entropy. I've had to toss a couple of aging switches in the skip recently for that.

If there's a mismatch it's going to be because the switch end is set half-duplex or autonegotiate.

I would be half inclined to set the Alpha briefly to autonegotiate with LANCP, and see if it succeeds. Or get someone to double check the equivalent configuration and statistics at the far end.

Re: %EWA0 possible duplex mode mismatch condition detected

Tim Schofield_1 — Wed, 14 Mar 2007 13:03:49 GMT

Thanks everyone for your help. Great forum. I will let you know how I make out.
BTW my second site with this issue had changed their switch to half duplex which caused the errors at that site.
Tim

Re: %EWA0 possible duplex mode mismatch condition detected

Ian Miller. — Wed, 14 Mar 2007 13:25:57 GMT

If in doubt blame the network - the network folks won't admit anything but the problem will mysteriously get fixed. :-)

I see this is your first question. Welcome to the forum. See
http://forums1.itrc.hp.com/service/forums/helptips.do?#33

Re: %EWA0 possible duplex mode mismatch condition detected

John Gillings — Wed, 14 Mar 2007 18:31:04 GMT

Tim,
Try this:

$ MCR LANCP SHOW DEVICE EWA0/INTERNAL

You will need PHY_IO privilege.

At the bottom you'll see "Driver Messages" which will give the state of the adapter. You may see something like:

12-FEB-2007 11:57:08.64 Auto-negotiation mode set by console (EWA0_MODE)

or

12-FEB-2007 11:57:08.72 FastFD mode set by console

If it's auto-negotiation, the messages should also say what was negotiated and when.

The rule is very simple. Both ends of the wire must be set the same.

Personally I recommend AUTO everywhere, as that's the default for most switches.

Re: %EWA0 possible duplex mode mismatch condition detected

rick jones — Wed, 14 Mar 2007 20:34:07 GMT

Some boilerplate on how autoneg is supposed to work and some indications when there is a duplex mismatch.

Re: %EWA0 possible duplex mode mismatch condition detected

Colin Butcher — Thu, 15 Mar 2007 02:44:13 GMT

In general the later NICs do autonegotiate of speed and duplex successfuly. LANCP is the tool you want to show the NIC states.

In general for EW 10/100 devices I prefer to set the speed and duplex at the console (eg: set ewa0_mode fastfd) and set the speed and duplex identically at the switch end as well.

For EW Gigabit devices (Broadcomm based I think) I let them autonegotiate as it then does a better job of link failure detection.

For EI devices I usually let them autonegotiate as the Intel 82559 based NICs seem to do autonegotiate better than the DEC SGEC, LANCE and other such chips used in the earlier Alpha NICs.

What is a real pain is if the site uses unmanaged switches and the Alpha NIC doesn't do autonegotiate well with that switch. I have that myself with my old PSW500au and a 3Com 10/100 switch, so I generally see what the switch thinks (look at the LEDs), then set the Alpha to the same state at console level. Usually works.

Cheers, Colin (www.xdelta.co.uk).