Coral outage this evening ...

Bill Murray bmurray at snf.stanford.edu
Sun Jun 1 12:41:17 PDT 2003


On Sat, 31 May 2003, John Shott wrote:

> 
> I'm worried that all of those mpathd errors on atu aren't healthy and I'm
> worried that the fact that those two actions "fixed" a significant network
> outage indicates that the problem is in our side of the world.  
> 

Not necessarily.  Mike made changes to our switch late Friday based on
Jason's explicit instuctions.  Since Mike did not make these changes 
permanently, the changes were undone when you rebooted the switch. 
I suspect that we do not have accurate information on integrating our
switch with the "CIS" network.  I have included Jason's email regarding
the changes below.

>From jconroy at cis.stanford.edu Sun Jun  1 12:36:02 2003
Date: Thu, 29 May 2003 10:11:12 -0700 (PDT)
From: Jason Conroy <jconroy at cis.stanford.edu>
To: John Shott <shott at snf.stanford.edu>, Mike Bell <mbell at snf.stanford.edu>,
     Bill Murray <bmurray at snf.stanford.edu>
Subject: spantree mode on the cisco switch

Some testing has suggested that the network connectivity problems all
share in common the new Cisco switch in the computer room. When set to
their factory defaults, Ciscos are known to be finnicky on mixed,
multi-vendor networks like ours.

In particular, we have trouble with the spanning tree algorithm. It's
designed to shut down a port when it detects a loop, such as when somebody
plugs a minihub into two wall jacks at the same time. However, it seems 
that many things can set it off. If you're switch's os is at all 
similar to mine, this has to be disabled on a per-vlan basis:

set spantree disable 1
set spantree disable 100
set spantree disable 980
.
.
.

Hopefully this will put an end to the sporadic packet loss that averaged
2% last night...

Jason


---------------------------------------------------------------------
 Jason Conroy                        Email: jconroy at cis.stanford.edu
 Systems/Network Manager                         Phone: 650 725-6115

        Center for Integrated Systems at Stanford University
---------------------------------------------------------------------




More information about the coral mailing list