Excessive Ethenet Collisions: v.091/2.0.36/21143

[Sami YOUSIF]

MJ Watson wrote:

> The network folks here have taken me off the hub, since the number of collisions
> coming out of my workstation is considered exessive...something like 40% of
> capacity (whatever that means).  Anyway, the only tulip option that works is
> the default, nothing else will connect with the network.  The hub is a 10Mbs
> half-duplex Cabletron.
>
> This is coming from another 2.0.36 workstation, which is running just under 20% collisions
> according to the network guy's stats.  I've included what I hope is some useful info
> below, though not being a network type I'm not sure what's really helpful.  I've tried
> (on the other system) eliminating any program that might send packets to the network,
> a list of running processes is included in case anyone might see an obvious culprit
> there.
>
> Any suggestions would be appreciated.

I do notice that it is in promisc mode....  Since you are running a sinffer
(netwatch)  You DO have permission from your net admin persons for promisc mode on that machine/subnet? (if not, and/or you did not switch it yourself, and
are not using a sniffer or arpwatch, then track down what/who is switching it to promisc mode {who is running netwatch?}).....

Also, check your network interface settings. They may be incorrect (see below),
and are causing  some problems... esp the broadcast on eth0

As for the daemons running on that machine, tcplogd does not need promisc mode
if logging  connections destined to the machine itself. same for icmplogd. xntpd  may be
configured for  broadcast mode, which does depend on the network interface settings for eth0 which seem to be incorrect.

Some cards when they start up with auto negotiation on, start up in 100mb mode.
when they cant  communicate, they then switch to 10mb mode.... On a 10mb hub, the 100mb mode before the card switches to 10mb may seem either like frame
errors, OR collisions
depending on the hub...

(ex: I have an intel netport Pro/100 printserver that is hooked to a 100mb
hub... when the print  server is first turned on, it causes 40-70% collsions for a few seconds and
then switches to 10mb  mode where it works fine)...

last thing is to check for cable length, and is it running near any flurocent lights?


> -----------------------------------------------------------------------------------------------------
> IFCONFIG:
>
> lo        Link encap:Local Loopback
>           inet addr:127.0.0.1  Bcast:127.255.255.255  Mask:255.255.240.0
>           UP BROADCAST LOOPBACK RUNNING  MTU:3584  Metric:1
>           RX packets:195 errors:0 dropped:0 overruns:0 frame:0
>           TX packets:195 errors:0 dropped:0 overruns:0 carrier:0
>           Collisions:0

This does seems strange... The loopback should show up as:

 inet addr:127.0.0.1  Bcast:127.255.255.255  Mask:255.0.0.0


>
> eth0      Link encap:Ethernet  HWaddr 00:00:F8:78:D2:8A
>           inet addr:10.40.42.209  Bcast:255.255.255.0  Mask:255.255.240.0
>           UP BROADCAST RUNNING PROMISC MULTICAST  MTU:1500  Metric:1
>           RX packets:15826553 errors:0 dropped:0 overruns:0 frame:0
>           TX packets:1789720 errors:0 dropped:0 overruns:0 carrier:0
>           Collisions:1152924
>           Interrupt:9 Base address:0xec00
>

This is not right.

check again the broadcast settings for eth0.. What you have does NOT seem
right.... AT MOST (default for class A 10.x.y.z net) it should be 10.255.255.255 with a mask of
255.0.0.0, but since it is subnetted, check with your network dudes/duddettes  what the correct mask and broadcast should be for your segment.

--
-
Sami Yousif
mailto:syousif@iname.com
http://www.mav.net/teddyr/syousif/ Personal Page

[eMail sent to any of my addresses is subject to the Conditions outlined
in https://www.mav.net/teddyr/emailtos.shtml]