First slide. Breathe. Now begins the fleshy monkey talky bits.
What you get after a FreeBSDpkg_add, a Redhat rpm –install, or apt-get, etc.
Extending graph coverage is easy. Go download a template, and if needed, drop a poller script in /usr/local/cacti/scripts/
Really. Collect –everything-You can never graph too much. You can overdue your Notification system, but you cannot overdo your historical trending data.
Sure. Ice Cream.Build sets of graphs from a dozen metrics, in about 20 seconds.
Migrating from the PHP Poller, to a compiled C poller gives a huge boost in performance, at the cost of a manageable amount of deployment effort.Install Boost, break the architecture up, and you can scale to quite well.
Aggregate graph creation plugin lets you get a more complete view of host clusters. If you have a more volatile environment and a static aggregate isn’t enough, alternate options exist allowing automated rebuild of aggregates.
Never deal with graph and tree creation again.
$ls –l /usr/local/cacti/cli/Most require you to run them via ‘php –q ./cli/script’.. All have –help. Read the docs.If you have a host database or a cmdb with an export function, or even a zone file with well named hostnames, and you can parse that in a script, wrapped around the cli tools, you can end up with a fully automated self-managing Cacti install. It’s a great place to be.
With nearly three years of operation in an extensively monitored, multi-colo, global environment with approx 2000 monitored servers; multi-poller Cacti can be accomplished with recent versions, and small-scale unofficial patches.
Current dev branches include integrated functionality for distributed polling.
Threshold plugin works like a Mini Nagios, only without the Nagios. Nectar emails off scheduled reports. Both are simple, powerful, and effective.
For me, Cacti works best when I can easily get data back out. Pull it out, feed it into a dashboard, use it for Nagios plugins, feed it into your business logic. Looking at your data is great, but parsing it and interpreting it relative to other data to get a complete view of your environment or application is priceless. Or, at least, many companies will charge you a very large price for it.
Interactive server utilization Heatmap, thanks to James Gladstone, exported Cacti data, and Google Charts.
Poller1’s running a little hot. It’s also on a slightly smaller box, and doesn’t get any new load from automatically added devices. If it was a VM, I’d consider reducing its allocated memory.
Participate in the forums, Read the docs, Contribute, Do Awesome Things.
Large Scale Cacti
Large Scale CactiAggregation, Automation, CMDB Integration, Extending, Reporting, and Scaling<br />Mike Lindsey – IT Architect - Cisco<br />
Basic Cacti<br /><ul><li> Requires a host with a web</li></ul>Server, a database, and PHP<br /><ul><li> Gives you a pretty interface, some core monitoring templates, access control, and clicky-clicky editing tools
Useful documentation provided online. Large and active community forums.
It’ll get the job done for a few racks of uninteresting servers, but quickly becomes limiting for complex, large, or volatile environments.</li></li></ul><li>Templates<br />A sometimes complicated process, that someone else has probably done for you.<br />Hit the forums. Good templates are stickied at the top of the “Scripts & Templates” forum.<br />If you want to graph it, there are probably templates already.<br />If there are no pre-existing templates, you might still not need to create them.<br />
Autom8Automatic Graph creation and Tree manipulation<br />
Command-Line Tools<br />.. Or, how to stop all the mouse-clicks, and make it work for you.<br />Add hosts<br />Add graphs to hosts<br />Add hosts to trees<br />Never manually add a host or graph again.<br />Or if CMDB integration isn’t your thing, check out the Discovery plugin!<br />http://www.cacti.net/downloads/docs/html/scripts.html<br />
Distributed Polling<br />Scale the bottleneck<br />Toss a poller vm in every colo?<br />