Boston Linux & Unix (BLU) Home | Calendar | Mail Lists | List Archives | Desktop SIG | Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings
Linux Cafe | Meeting Notes | Blog | Linux Links | Bling | About BLU

BLU Discuss list archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[OT] Labeling wires



Inspired by the nature of the business where I was working (up until *very*
recently), I came up with the concept of a "rainbow flag" cable-coloring
scheme to replace past failed cable-labeling strategies.

What I did was assign a primary color to each Ethernet switch, and (other than
the uplinks) *all* cables connected into each switch are all the same color.

Most of the servers had 5 connections:  four on the motherboard (all the
latest servers from Dell/HP have 4 RJ45 jacks) plus a KVM for the out-of-band
management (DRAC, ILO, or whatever it is Super Micro calls theirs).

One other major change about my latest design was that I discarded patch
panels.  Servers are directly connected to the switch, without going through a
patch panel.

By doing this, I eliminated practically all the labor costs of tracing cables.
 You *know* which switch each server port is connected to, based on proximity
and color coding.

My color coding was this:  internal LAN was red and blue (red=primary,
blue=alternate -- Ethernet channel bonding), DMZ was orange and purple (again,
a primary and backup), console current-generation out-of-band net was yellow,
old-fashioned (Raritan) out-of-band KVM network was green, and inter-switch
connections were black and white.

With 200 servers in a cluster you can easily have 1500 cables.  You'll wind up
with twice that many if you use patch panels (vs. mounting the Ethernet
switches in racks near the servers).  My color-coding method reduced
maintenance effort considerably, and eliminated all but one failure mode: 
once in a while someone would cross-link two same-color switches on a port
other than the designated uplinks.  If you don't lock down your spanning-tree
configuration, such cross-linking will create a nasty packet-storm loop.

The nicest thing about this approach was the training.  It's really easy to
train a new tech how to rack a server by this color-coding strategy, which can
also be used on the power side (three colors for the 3 phases of utility
power, and two other colors for A-side and B-side UPS power--alas the power
cables all come in black, vs. Ethernet cables which come in about 10 different
colors, so you still have to use tape to do the power color-coding.)

-rich







BLU is a member of BostonUserGroups
BLU is a member of BostonUserGroups
We also thank MIT for the use of their facilities.

Valid HTML 4.01! Valid CSS!



Boston Linux & Unix / webmaster@blu.org