Veritas-bu

[Veritas-bu] Is there an official document of client and serv

2004-03-18 12:59:27
Subject: [Veritas-bu] Is there an official document of client and serv
From: ddunham AT taos DOT com (Darren Dunham)
Date: Thu, 18 Mar 2004 09:59:27 -0800 (PST)
> We've found that the switch can be automatic while the server is forced to a
> setting and all will be well. 
> 
> This lets the network infrastructure be easily maintained while individual
> SA's can force their servers to whatever settings they want.

While such a configuration can work, I would advise against it.

In many cases, "forcing" an interface will disable the auto-negotiation
protocol, so that the partner interface cannot successfully determine
what it should do.  I would recommend setting both sides or neither.

In any case, monitoring interfaces for errors is the easiest way to
check for problems.  For the most part, I let things default to
auto-negotiated unless a problem appears.  Whenever a split-duplex
situation happens, you can see errors on the interface in 'netstat -i'
output.  Monitoring should pick that up immediately.

If the host is full and the partner is half, the partner will be sending
"jam" signals as it detects the simultaneous packets as "collisions".
These jams will be interpreted on the local host as input errors.  If
the opposite has occured, you will see collisions and output errors on
the host.

For me, anything above about 0.01% error rate merits investigation.
Split duplex with much traffic will send the error rate high (often
above 1%).

-- 
Darren Dunham                                           ddunham AT taos DOT com
Senior Technical Consultant         TAOS            http://www.taos.com/
Got some Dr Pepper?                           San Francisco, CA bay area
         < This line left intentionally blank to confuse you. >