input packet errors

[Sunny Jaisinghani]

Hi Gurus,
We recieve lots of input packet errors. we clear it using lanadmin.

Is there a way to avoid this?
A patch which can take care of this?

we don't have access to the hardware.

any help will be appreciated.

thanks
sunny

[Peter Godron]

Sunny,
is this down to the machine or a noisy line ?
You could try and lower the connection speed ?

[Sunny Jaisinghani]

Hi Peter,

is this down to the machine or a noisy line ?

---> yes it is a noisy line. it is a production server.

You could try and lower the connection speed ?

--> can you be more clear with the above point. which parameter should i tweak?
the baud rate.
packet size.

Since i am new to HP-UX i haven't done much research due to limited resources.

thanks

[linuxfan]

Hi Sunny,

Is the lan card configured to auto-negotiate or is set to 100FD?

Get the PPA from lanscan (look for Card In#)

then run lanadmin -x PPA#
Example: lanadmin -x 0

If there's a mismatch in speed/duplex settings on the server and the switch side it could result in lots of packet errors.

To correct the issue use "lanadmin -X 100FD 0" and update the appropriate file in /etc/rc.config.d/

To figure out which file to modify, run /usr/sam/lbin/laninfo and that'll tell you the btlan or gelan etc. Now modify the /etc/rc.config.d/100btlanconf or the appropriate file to ensure the lan settings are configured in subsequent reboots.

[Sunny Jaisinghani]

thanks a lot

i'll make the above checks and update the post asap.

Regards

[rick jones · Accepted solution]

Before you go hard-coding anything, make sure the statistics from lanadmin suggest a duplex mismatch, and consider making both sides autoneg if there is a duplex mismatch. See the attached.

[Sunny Jaisinghani]

Hi

I have got the following o/p from lanscan and lanadmin command.
============================================
[xxxxxx:/]# lanscan
Hardware Station Crd Hdw Net-Interface NM MAC HP-DLPI DLPI
Path Address In# State NamePPA ID Type Support Mjr#
0/0/0/0 0x001083FF1AA0 0 UP lan0 snap0 1 ETHER Yes 119
0/3/0/0/4/0 0x00306E2564D0 1 UP lan1 snap1 2 ETHER Yes 119
0/3/0/0/5/0 0x00306E2564D1 2 UP lan2 snap2 3 ETHER Yes 119
0/3/0/0/6/0 0x00306E2564D2 3 UP lan3 snap3 4 ETHER Yes 119
0/3/0/0/7/0 0x00306E2564D3 4 UP lan4 snap4 5 ETHER Yes 119
===========================================

[xxxxxx:/]# lanadmin -x 0
Current Speed = 100 Full-Duplex Auto-Negotiation-OFF
[xxxxxx:/]# lanadmin -x 1
Current Speed = 100 Full-Duplex Auto-Negotiation-OFF
[xxxxxx:/]# lanadmin -x 2
Current Speed = 100 Full-Duplex Auto-Negotiation-OFF
[xxxxxx:/]# lanadmin -x 3
Current Speed = 10 Half-Duplex Auto-Negotiation-ON
[xxxxxx:/]# lanadmin -x 4
Current Speed = 10 Half-Duplex Auto-Negotiation-ON

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

The above statistics are from the server side. i don't have access to the hardware. how can i find switch statistics remotely.

Rick:
from the above o/p it is not a 1000Base-T n/w.
And i think there is a duplex mismatch(sorrow) :)
However before taking any action i want to confirm the settings on the switch side.

Any suggestions ?

Thanks and Regards

[rick jones]

If you side is hard-coded full duplex and you see the FCS errors, if there is a mismatch with the switch it means that the switch is in half-duplex. It got there either via "failed" (bad term but best I have at the moment) autonegotiation, or because it was hard-coded to half-duplex.

Per the happiness/lucky/sorrow table, that suggests (to me anyway :) that you should try using auto. If the switch is auto, all should be well. If the switch is hardcoded to half, you will be in the "lucky" space.

If the switch is a "managed" switch, sometimes you can retrieve its stats via SNMP. I believe that HP ProCurve switches have various management software you can run on a PC as well. Both might require knowledge of a magic password set on the switch though.

[Sunny Jaisinghani]

Hello Rick,

Thank you for Suggestions.

I have few more doubts.

1. How can i know which nic is responsible for the packet errors.
There are 5 nics with different speed/duplex settings. Some of them are auto some are not.
Shall i assume that the packet errors are due to nics which have autonegotiate OFF.????

2. If the switch is a "managed" switch, sometimes you can retrieve its stats via SNMP. I believe that HP ProCurve switches have various management software you can run on a PC as well. Both might require knowledge of a magic password set on the switch though.

Is there are any link where i can read some info about the point mentioned above.

Regards
Sunny

[rick jones]

lanadmin -g mibstats gives per-interface statistics. so you already know which interface when you run lanadmin.

as for the switch stuff, start at http://www.hp.com/go/procurve for HP ProCurve switches and their manuals etc.

[Sunny Jaisinghani]

Hello Rick,

Thanks for your help.

Regards

[Sunny Jaisinghani]

Hi,

this is regards with the above problem.

1. what will be the consequences if i don't clear the LAN ineterface statistics registers. ????

2. is there a limit on these registers which store the count of input packet errors.????

Thanks in advance
Sunny

[rick jones]

They are only statistics, clearing the stats does nothing other than clear the stats. IIRC until 11iv3 (11.31) they are all 32bit counters. And if the error stats happen to wrap, frankly that is no big deal either.

[Sunny Jaisinghani]

Thanks Rick,

u sure if the value of the statistics wrap around the system won't throw any error.??

or it won't be hazardous to the system.??

[rick jones]

Wrapping of statistics counters in networking is not something you need worry about, so do you not need to worry about clearing input packet error statistics.

[Sunny Jaisinghani]

ok Rick

Thank you for sharing your knowledge