A few days ago Hackaday featured an article about my Raspberry Pi cluster: http://hackaday.com/2013/11/05/can-an-8-node-raspberry-pi-cluster-web-server-survive-hackaday/
The article was published at 12.00GMT, and my cluster served 2,440 pages in the first hour. In 12 hours, my site served 21,000 pages.
I checked the load on the load balancer and the backend servers, and everything seemed to be fine. This is the most traffic my site has had to handle so far, and I was pleased with the way it went. Page load times stayed low, and the cluster still had plenty of spare capacity.
Feedback was generally good. Some people questioned whether I really need an eight node cluster. I guess I don’t need eight Pi servers, especially for the amount of traffic my site is currently handling.
There are some dynamically generated pages on my site, and serving these pages uses a disproportionate amount of CPU time. I will probably add more dynamic content in the future, so I believe it’s a good idea to have some spare processing power.
It’s also important to consider that visitors don’t arrive on my site at evenly spaced intervals. If someone tweets a link to my site, there may be a rush of visitors in a short space of time, so the cluster needs to be able to handle spikes in traffic.
Some people left comments suggesting that my internet connection could be a bottleneck, and my internet connection will choke before the cluster gets overloaded. This seems like a reasonable point. If this becomes a problem, I may need to upgrade my internet connection.
Over the last few days I’ve been working on setting up uWSGI with Nginx. Everyone says Nginx is much faster than Apache when serving static content. Most of the pages on my site are cached, so it will be interesting to see how well Nginx performs with the CMS that I’ve written. I’ll write an article about Nginx and uWSGI in a few days.