Interface Errors

[mshields at namemedia.com (Matt Shields)]

Any cheap switch from CompUSA couldn't handle the traffic we pass.
60-80Mbits constant outbound, 15-20Mbits inbound.  This isn't burting
traffic either, this is sustained traffic 24x7.

Also, if the Dell switches were the problem.  Wouldn't the switches show
interface errors as well?  We only see errors being reported in the OS
(RHEL 3)


Matthew Shields
Sr Systems Administrator
NameMedia, Inc.
(P) 781-839-2828
mshields at namemedia.com
http://www.namemedia.com


-----Original Message-----
From: David Backeberg [mailto:dave at math.mit.edu] 
Sent: Tuesday, June 20, 2006 9:12 AM
To: Matt Shields
Cc: discuss at blu.org
Subject: Re: Interface Errors

I've heard that Dell switches are absolute pieces of garbage, and that
you get exactly what you pay for when you buy them. Do you have any
non-Dell switches you can put in place to see if the interface errors
will go away then? You can get little five-port GigE switches from
Netgear for about $50 these days. I realize I'm recommending a crap
switch to isolate what was more expensive hardware, but you haven't
ruled out your switch as a problem yet. Just swapping a Dell switch with
another Dell switch doesn't rule out that their hardware can't handle
your load.

http://www.webhostingtalk.com/archive/thread/437041-1.html
gs "dell ethernet switches crap" for more entries like that

If everything in your datacenter is Dell, you can put a non-Dell switch
between your server and the next switch, and see if the error gets
isolated by the intermediate non-Dell switch instead of getting passed
on to the server. You could accomplish the same thing with a dual-nic
linux bridge box. That way, you could use ifconfig / iptables / ebtables
to monitor both sides of the connection and isolate errors to one side.

Dell server <> bridge box eth0 <> bridge <> bridge box eth1 <> Dell
switch http://linux-net.osdl.org/index.php/Bridge

I'm also assuming you haven't done anything idiotic, like run your
ethernet cables on the floor and had given tours to people in
high-heels, or run your ethernet cables along flourescent lights, or
quick-tie the ethernet cables to high-current power cables, and wonder
why you're getting errors.

-- 
David Backeberg (dave at math.mit.edu)

On Mon, 19 Jun 2006, Matt Shields wrote:

> I've been having a problem for about a week where the interface on my
> server slowly collects errors (~4 or 5 errors/minute). This server
(Dell
> Poweredge 1850) is our firewall and is connected via GigE to our Dell
> PowerConnect 5324.  The Powerconnect has an uplink to our provider
that
> is Gig Fiber.  I'm using CAT6 cable and I have replaced it numerous
> times.  I've also changed the switch port.  Can anyone offer any
> suggestions?  Don't know if this matters, but we're passing 60-80
Mbits
> outbound, 15-20 MBits inbound.  
> 
> 
> Also, I'm monitoring the errors with ifconfig, not SNMP.  NET-SNMP
seems
> to have a problem with 64-bit counters, which we need since we pass
such
> a large amount of traffic.
> 
> 
> Any ideas?
> 
> 
> Matthew Shields
> Sr Systems Administrator
> NameMedia, Inc.
> (P) 781-839-2828
> mshields at namemedia.com
> http://www.namemedia.com
> 
> _______________________________________________
> Discuss mailing list
> Discuss at blu.org
> http://olduvai.blu.org/mailman/listinfo/discuss
>