I have an operations background. My first company taught me most of what I know about how to run software and server operations. Fast-forward 15 years and we are now all about the cloud, VPSs, and Kubernetes. I love the cloud. Up until a few weeks ago, my blog has been hosted at Scaleway, which has worked great for me. Today I run it on my own server where (for better or for worse) everything is managed by me.
One thing I was not happy about was Google Analytics. To keep my uptime I want to know the number of page loads and system load in order to optimize and scale. I know, I should probably be using a CDN to mitigate some of these issues, but I don't feel I'm there just yet. Google Analytics is one of those services that is not known to be privacy friendly, and if you are here, I respect you and your time. I don't include ads and I try to keep the tracking as limited as possible disabling social crap, etc. For my purposes, I don't need Google analytics. A web server logs all of the information I need for scaling purposes. All I needed was to access those logs (which I already had access to) and store the data in a database, create a dashboard, and kiss Google Analytics goodbye.
I know, I could've used AWS or Google Cloud to do this; but the cost over time would have been prohibitive. Self-hosting seems like the right answer at the moment.
The game plan:
With the help of my company, I got a new server and some data center space (power, networking, and a rack). I know, this is the most tricky part as not everyone works for a telco that can provide these things. The point of this post is not to justify the financial advantage of self-hosting vs the cloud, but to point out the elements we overlook by leaving it up to the cloud to do some of the heavy lifting.
I installed ESXi on the server to run all my infrastructure. I have done this before, so I felt fairly comfortable reproducing this.
I used VyOS for all the networking and firewall needs. This was the trickiest part. I hate networking. I still do and the networking concepts, to be honest, just beat me. Somehow though, with basic subnetting and routing skills, you can actually get surprisingly far.
NGINX to host my site (quite straight forward).
Run an Elastic stack (really, just Beats, Elasticsearch, and Kibana) for data processing. From system auditing, to security, log parsing, and metrics. This stack is the central unit that gives me visibility into what's happening inside my system. This includes NGINX log analysis.
Over the next few weeks I'll be writing about my experience moving away from the cloud. The work it involved, where I believe it's better than the cloud, and where I believe the cloud is superior. I will talk about what's left in my set up and how I'm planning on tackling it.
They say the journey is as important as the destination itself and, in this case, I must agree. I have learned a lot through the process. Perhaps someone will learn something from my experience. That will make it all worth it!