Part V The Final Product

The Final Product

As each web server runs the metric module parsing each access-log, a set of 5 metrics are sent to my collector. Every 60 seconds that set may change with differently named metrics. It's ok though because DynamicGraph will age out all but the top5 graphs, building the .json and keeping my metrics appearing current.

This gives some of my Ganglia graphs a more "UNIX Top" like feel and increases the possibilities of what can be done with Ganglia - both as a trending platform and as a current real-time dashboard. In my example here I am able to check at any time on any node to see who might be overloading us or just get an idea of volume.

I have also applied this in similar cases to track current swap activity, current I/O reads/writes, and current memory usage.

So, while it may be a bit "hacky" to run dynamic metrics that provide LINUX "top-like" views, it is definitely possible and works well.

Below one can see an example metric. Here I have blanked out the customer names but as you can see the dynamic nature of the metric works. Every 60 seconds the customer names refresh depending upon volume and the values tell me how many requests per customer occurred over the past 60 seconds. The blue and red lines extend further back in history because those two customers always show up in the top5 - they are very busy. The green violet and yellow lines appear to the right side of the graph because they recently refreshed in the past run.