[mrtg] Re: MRTG management
bwoodcoc at genuity.com
Tue Nov 26 04:09:25 MET 2002
Hmm, this may take a bit of typing. Our scope is 4-500 switches and about
120 routers. Total interconnections are around 800-1000 interfaces for
collection is my best guess (but no stats on user ports). We are using mrtg
with rrd's and 14all.
A cfgmaker like process is run daily to keep all configs up to date and
fairly dynamic (run on the fly if needed). Configs are broken into 4
directories for routers, switches, networks (dummy configs for navigation),
and interfaces which is on the fly dummy configs also.
The rrd's are stored as such under apache htdocs: L3/<rtrName>/*.rrd and
also L2/<sw_ip>/*.rrd . We are keeping utilization and ifErrors/ifDiscards
on every interconnectivity interface. We also collect response time
(*distributed*) for across every point to point link (some things I think
you still can't buy). Thresholds on everything with alarms going to a
central web based alarm manager.
Navigation comes in multiple forms.
1. A small cgi (so it's dynamic) lists out either all the routers or all
the switches. These lists are text hot linked to a bridge script which
calls 14all with a pointer to the cfg and the rrd. Simple, reliable, but
not flashy at all.
2. The alarms on the web page have a hot link to the cgi which calls 14all.
So we have an alarm manager displaying any issue and one click to view
metrics for that condition.
3. Metrics are exported to csv files every 5 minutes. A topN script can be
run which can sort this stuff multiple ways and there are hot links which
again point to the cgi which calls 14all from the listing.
4. I have scripts which auto-maps all this topology and use this as
graphical navigation to mrtg data. It looks like a visio topology drawing
of sorts and for cisco running native ios it is an integrated Layer2/Layer3
drawing. However it is not visio, they are drawn on the fly using gd/GD.pm.
The maps are png based with all objects (lines, rtrs, switches) as hot
links back to the cgi which calls 14all. There is also submap definition
and navigation. Also the lines are color coded by either speed, exported
utilization values, or exported response values (on the fly when requesting
the page). Very slick.
This environment for troubleshooting and analysis is nothing short of
*superb* imho (daddy is proud of his kids along with mrtg/14all). However
it took a great deal of effort and much time. But hey......you asked ;-).
Someday I'd like to offer this to the public and try this stuff out if they
wish. I've made some strides towards that end, however at this time I'm not
there yet. Hmm, maybe I should more formally describe the scripts in a doc
for folks to look at. For those who may be interested in such a doc
describing this environment in more detail (and its limits) drop me an
email. If there is significant interest I will do this and get it onto the
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
FAQ http://faq.mrtg.org Homepage http://www.mrtg.org
More information about the mrtg