Prev: Server crash
Next: Can't access website from Linux
From: Elliot J on 8 Nov 2007 07:58
Hello,

If I run ifconfig on one of our servers, a quantity of collisions are
reported on one of the NICs. Should I be concerned?

eth2 Link encap:Ethernet HWaddr 00:0A:5E:58:37:EB
inet addr:10.11.1.1 Bcast:10.11.1.255 Mask:255.255.255.0
inet6 addr: fe80::20a:5eff:fe58:37eb/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:23921568 errors:0 dropped:0 overruns:333 frame:0
TX packets:20756867 errors:0 dropped:0 overruns:0 carrier:44
collisions:1185571 txqueuelen:1000
RX bytes:3487241823 (3325.6 Mb) TX bytes:1760227159 (1678.6 Mb)
Interrupt:10 Base address:0x5000

Thanks.


From: Robert Harris on 8 Nov 2007 08:35
Elliot J wrote:
> Hello,
>
> If I run ifconfig on one of our servers, a quantity of collisions are
> reported on one of the NICs. Should I be concerned?

No. But your network would be faster if you replace your hub with a switch.

Robert

>
> eth2 Link encap:Ethernet HWaddr 00:0A:5E:58:37:EB
> inet addr:10.11.1.1 Bcast:10.11.1.255 Mask:255.255.255.0
> inet6 addr: fe80::20a:5eff:fe58:37eb/64 Scope:Link
> UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
> RX packets:23921568 errors:0 dropped:0 overruns:333 frame:0
> TX packets:20756867 errors:0 dropped:0 overruns:0 carrier:44
> collisions:1185571 txqueuelen:1000
> RX bytes:3487241823 (3325.6 Mb) TX bytes:1760227159 (1678.6 Mb)
> Interrupt:10 Base address:0x5000
>
> Thanks.
>
>
From: Geoffrey Clements on 8 Nov 2007 08:54

"Elliot J" <elliotj(a)x-x.co.uk> wrote in message
news:fgv16c$t6h$1$8300dec7(a)news.demon.co.uk...
> Hello,
>
> If I run ifconfig on one of our servers, a quantity of collisions are
> reported on one of the NICs. Should I be concerned?
>
> eth2 Link encap:Ethernet HWaddr 00:0A:5E:58:37:EB
> inet addr:10.11.1.1 Bcast:10.11.1.255 Mask:255.255.255.0
> inet6 addr: fe80::20a:5eff:fe58:37eb/64 Scope:Link
> UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
> RX packets:23921568 errors:0 dropped:0 overruns:333 frame:0
> TX packets:20756867 errors:0 dropped:0 overruns:0 carrier:44
> collisions:1185571 txqueuelen:1000
> RX bytes:3487241823 (3325.6 Mb) TX bytes:1760227159 (1678.6 Mb)
> Interrupt:10 Base address:0x5000
>

Possible network congestion or you need to check the duplex settings at both
end of the physical link.

--
Geoff


From: Moe Trin on 8 Nov 2007 14:56
On Thu, 8 Nov 2007, in the Usenet newsgroup uk.comp.os.linux, in article
<fgv16c$t6h$1$8300dec7(a)news.demon.co.uk>, Elliot J wrote:

>If I run ifconfig on one of our servers, a quantity of collisions are
>reported on one of the NICs. Should I be concerned?
>
>eth2 Link encap:Ethernet HWaddr 00:0A:5E:58:37:EB

[compton ~]$ etherwhois 00:0A:5E
00-0A-5E (hex) 3COM Corporation
000A5E (base 16) 3COM Corporation
5400 Bayfront Plaza
M/S 5119
Santa Clara CA 95052-8145
UNITED STATES
[compton ~]$

3Com - no indication which model

> inet6 addr: fe80::20a:5eff:fe58:37eb/64 Scope:Link

Do you know that IPv6 is available on this link?

> UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
> RX packets:23921568 errors:0 dropped:0 overruns:333 frame:0
> TX packets:20756867 errors:0 dropped:0 overruns:0 carrier:44
> collisions:1185571 txqueuelen:1000

[compton ~]$ echo 1185571/20756867 | bc -l
..05711704950462899820
[compton ~]$

5.7 percent? For a half-duplex link using coax or a hub, that's not
unreasonable. The fact that you've got a few input overruns suggests
that the link and this computer are "busy". The 44 carrier errors
while low are a slight concern (shouldn't be any - perhaps the cable
got disconnected momentarily).

Old guy
From: Peter Benie on 8 Nov 2007 15:02
In article <4733113f$1_1(a)glkas0286.greenlnk.net>,
Geoffrey Clements <geoffrey.clementsNO(a)SPAMbaesystems.com> wrote:
>"Elliot J" <elliotj(a)x-x.co.uk> wrote in message
>> If I run ifconfig on one of our servers, a quantity of collisions are
>> reported on one of the NICs. Should I be concerned?
>>
>> RX packets:23921568 errors:0 dropped:0 overruns:333 frame:0
>> TX packets:20756867 errors:0 dropped:0 overruns:0 carrier:44
>> collisions:1185571 txqueuelen:1000
>
>Possible network congestion or you need to check the duplex settings at both
>end of the physical link.

The interesting part is overruns and carrier errors, which should be
vanishingly rare. These indicate that your network is not performing
within spec, ie. it's not merely due to network congestion.

I agree with checking that the duplex settings match. The symptoms of
this are that small packets work fine and large packets be very, very
slow. If your equipment won't tell you what duplex settings are in
use, you can check this by sending pings of various sizes.

If that's not the problem, other things to check are:
* network diameter, esp. if you have daisychained 100Mb/s hubs.
* cable faults, eg. badly crimped UTP or mismatched pairs.

I've seen switches go wrong when connected to certain NICs, but this
is very rare. If this happens, you can put a mini-switch between the
affected NIC and the building wiring. Sometimes the problem is
specific to certain ports on the switch and there may be upgrade to
the switch firmware to fix it.

Peter
 | 
Pages: 1
Prev: Server crash
Next: Can't access website from Linux