LAN interface configuration and Error messages

[Yogeeraj_1]

Hello experts,

I think the LAN interfaces have not been properly configured on my L1000.

It may be a Duplex issue that we do not master completely....

When i check the status (speed) of all my lan cards, i get the following output:
============================================================
L1000: home/deg>./speed-lan.sh
Card at PPA 0 - IP Address: 132.147.160.9 - Speed = 100 Full-Duplex Auto-Negot
iation-OFF
Card at PPA 1 - IP Address: Not assigned - Speed = 100 Half-Duplex
Card at PPA 2 - IP Address: Not assigned - Speed = 100 Half-Duplex
Card at PPA 3 - IP Address: Not assigned - Speed = 100 Half-Duplex
Card at PPA 4 - IP Address: Not assigned - Speed = 100 Half-Duplex
Card at PPA 5 - IP Address: 10.0.0.100 -
L1000: home/deg>
============================================================
My HP J4812A ProCurve Switch 2512 ports are configured as follows:
============================================================
Port Type: 10/100TX
Enabled: Yes
Link Status: UP
Current Mode: 100FDx
Flow Ctrl: OFF
Bcast Limit: 0
============================================================


And it is giving me the following errors:
============================================================

Excessive CRC/Alignment Errors on port 2
23-Aug-02 10:42:14 AM

Description:
A high percentage of data errors was detected on port 2.

Possible causes:
The possible causes include faulty cabling or topology, half/full duplex mismatch, a misconfigured NIC, or a malfunctioning NIC, NIC driver, or transceiver.

Actions:
If port 2 is 100Base-T, make sure the cable connectors, punch-down blocks, and patch panels connecting to that port are Category 5 or better. Verify the correctness of the installation using a Category 5 test device.
Check the directly-connected device for mismatches in half/full duplex operation (half duplex on the switch and full duplex on the connected device, or the reverse).
Update the NIC driver software.
Verify that the network topology conforms to IEEE 802.3 standards.
Replace or relocate the cable. Also check the wiring closet components, transceivers, and NICs for proper operation.


============================================================

I want to know:
1. Where we have done the wrong configuration (if any)?
2. When to use Full and when Half Duplex?
3. Can the settings be changed on-the-fly or do i need to bounce the server after making the change?
4. How can i check if my LAN interface settings have been configured for optimal performance?

Thank you in advance for your replies and precious guidances.

Best Regards
Yogeeraj

[U.SivaKumar_2]

Hi,
Hope this solve the issue.
Flow Ctrl: ON

regards,
U.SivaKumar

[sven verhaegen]

Ok

I don't even have to think about this problem ,it's been happening so much that I can almost predict the cause , this surely is a DUPLEX mismatch issue , the execessive CRC error visible on your switch (or HP) ar caused by the constant abnormal packet termination/collision cased but a full duplex to half duplex setup , checking further on your switch data you see FD set for the port mode .. let me explain the rules for negotiating duplex

HP switch state

full full ok
half half ok
negot. negot. ok

half negot. nok
full negot. nok
half full nok
full half nok

you have to have both sides set tot he same mode , further if one side is negotiating an the other isn't that also can fail as negotiation need both partners to negotiate otherwise no data is exchange and it depends on the verdors card settings what a specific port will end up in ...

I advise you to ewither set the interfaces on the hp-ux autonegotiating as well as the prots of the switdch or better to force both to FD , on the switch this is apparently the case 9just check to make sure) on the HP-ux , adapt the interface file controiling these setting , the file is dependant on the type of lancard used and can be found in /etc/rc.config.d , the file alwasy resemble sosmething like hp...conf , for instance a hp-pb card would use hp100btconf , in there there are parameters for the landriver name , a posible mac address change and a speed and duplex seting , put the duplex setting at 100FD and this will solve the issue

.. i don't know how you check on the HP-ux the lancard setting but the best way is to use lanadmin , run the command and in the lan menu displaying the data of your lancard there will be a line describing the card isself and at the end of that line the duplex & speed setting of your lancard can be found , you can manually adapt the speed and duplex using the command line option of lanadmin , lanadmin -x will actively get the staus of the card , lanadmin -X 100FD will set the speed to forced 100 fullduplex , the only effect this has is to reset the interface , it doesn't break the connection on TCP for instance , you only loose connection tot he system for 2-3 sec period


using full or half duplex just depend on the need , the fatc is that the only basic differensy be the way you transfer the actual data on the line , either by block sending on the complete whire and waiting for return data (half) risking collisions or by separating the upper and lower pair for daat send and receive effectively discarting any collision risk , howver the speed gain is almost reduced by the fact the bandwith in full(mode) is reduced so performance only differes between the two modes in heavily loaded networks where lots of collision exist (e.g. flat network with hubs)

[Joaquin Gil de Vergara]

how is the conection betwen 9000 and Procurve?

Is direct conection or you have any hub?

[sven verhaegen]

good remark on the flow control , hope the data was't to much or not clear if any questions remain just let us know

[Ron Kinner]

Yogeeraj,

I guess since there is no rabbit that you are still having a problem.

First, are you sure that port 2 on your switch goes to this HP box? (I'd disconnect it and verify that the link light on the HP goes out or that lanadmin
lan
display
shows it down.)

Second, does it go to PPA 0? (The one with the 132.147.160.9 IP address.)

If it does there is no sign of a duplex mismatch since both seem to say they are 100 full.

Could we see both pages of
lanadmin
lan
display
from your HP?

Have you tried:

Using a different port on the switch?

Using a different NIC on the HP?

Replacing your cables?

Clearing the error counters on your switch and verifying that the problem still exists?

Ron

[harry d brown jr]

Yogeeraj,

What is plugged into port 2 of your procurve??

live free or die
harry

[Yogeeraj_1]

Hello everybody,

Sorry for this late feedback, I still need to have to go physically check my connections on my HP switch. I will be at the site (where we keep our servers) on Monday morning.

I will then reply.

I have already done some tests using telnet, but need to cross-check before i can give you some valid feedbacks.

Thank you all for your time and replies.

Best Regards
Yogeeraj

[rick jones]

I disagree slightly with the table for auto versus hardcode. When one side is set to half, and the other auto, the auto side will fail. autoneg failure requires falling-back to half duplex. so, half/auto should be OK, but not for a good reason :)

With my HP NICs connected to procurve switches, I've never had any problem with autonegotiation, so I've never hardcoded any of my NIC or port settings. Perhaps my life has been charmed.