My very tardy summary:
Summary of suggested ideas to fix Oerr count with an hme interface.
I have swaped lots of wires, some switches, applied all applicable
patches, recieved a new network card, and am still showing the same
problems. I have found out that SUN suggests that you not autonegiate with
hme cards, though. I am currently in the process of getting SUN support
to tell me what babble and tx_late_collosions are.
Matt Goebel
---- >From john@alpo.casc.com.casc.com Wed Oct 21 14:37:31 1998You don't say, but did you first check all the cables on your system and possibly replace them all with new cables? Have you tried moving the system to a new port on the switch to make sure it isn't a problem on that end? From the sound of it, that's most likely where the problem is. ---- >From mark_conroy@em.fcnbd.com Wed Oct 21 14:49:19 1998
I think that in normal operations, the acceptable number of output errors should be 0.025% of the total number of output packets. You will see a higher number of errors as a result of pulling cables and working on the interface card. Does the number below for (Oerrs-2273) also reflect pulling the cable and shutting down the interface. That may have added to the output errors. ---- >From pete@cobra.brass.com Wed Oct 21 15:14:04 1998
Make sure that the machine is booted with the external hub connected. This is very important because the 'hme0' interface auto-negotiates with the hub for the speed and duplex mode. This does not occur when the machine is booted without the hub connection and the machine and hub very often will default to incompatible modes.
You should also send the hme0 results from 'netstat -k' as this will provide additional insight into the types of errors that are occurring. ---- >From jsturges@gmri.com Wed Oct 21 15:56:57 1998
Hello, actually, I think _input_ errors are more indicative of bad hardware, not so much _output_ errors. Output errors usually indicate network trouble. Basically, if you have output errors, and you're on a shared ethernet segment, it may well be as simple as an overloaded segment. If your Oerrs is in the 5 - 10% range of all output packets, that's probably a good sign of network overloading. Of course, you need a Sniffer or something similar to really be sure. If you're on a switch, then another common cause for high error rates like that can be a _duplex_ mismatch between the host and the switch port. Use "ndd -get /dev/hme <parameter>" to check the interface; ndd will return a 0 or 1. Make sure the returned values match what the switch is set at. Here's a table for <parameter> values:
parameter values -------------- -------- link_status 0=down, 1=up link_speed 0=10Mbps, 1=100Mbps link_mode 0=half-duplex, 1=full-duplex.
Further, since you're running 2.5.1, make sure you've got the latest hme driver patch (patch no. 104212-13). This is particularly important if you're running at 100Mbps.
good luck, Jonathan ---- >From marksh@funb.com Wed Oct 28 07:48:46 1998
You should try checking the link mode and speed to be certain it is compatabile with the your network infrastructure. I had this problem and we set our switch to full 100, but still got errors until i edited the /etc/system and rebooted.
ndd /dev/hme link_mode (0=half-duplex, 1=full duplex) ndd /dev/hme link_speed (0=10M, 1=100M)
to force 100full duplex,
/etc/system
set hme:hme_adv_autoneg_cap=0 set hme:hme_adv_100hdx_cap=0 set hme:hme_adv_100fdx_cap=1
-- Matt J. l. Goebel : goebel@emunix.emich.edu : LT staff @ EMU : Hail Eris Neo-Student, Net Lurker, Donut consumer, and procrastinating Furry Fan. "...When we were living in a dream world, Clouds got in the way We gave it up in a moment of madness, And threw it all away..." E.W. & A.P.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:51 CDT