Jekyll and Amazon AWS hosting
— Initially published on 10 Apr 2015
I used to host my website on a virtual private server. Maintaining a server is always a useful skill to have, especially when using a “infrastructure as code” approach. See my article on Vagrant as an example.
My VPS instance used to be hosted by Gandi. I have been a Gandi customer for 15 years to deal with domain names, and I jumped to them for hosting when they ventured into the IaaS offering business.
I can only say positive things on the Gandi IaaS offering. It has always been rock solid in my experience. It is a bit pricey, but you get what you pay for.
As I wasn’t using the server instance for anything more than running a HTTP server for static files, I looked into a cheaper option (I was paying around 9 euros per month for a tuned-down VPS).
I could have gone the GitHub Pages way, but I prefer the flexibility of custom Jekyll plugins, or the ability to switch to another static site generator. I also wanted to keep the HTTPS support.
So I went with Amazon AWS.
Tools of the trade
I generate my websites with Jekyll, mostly because it is as minimalistic as it needs to be. Jekyll just gets out of the way with minimal fuss.
The produced static website is then stored to a S3 bucket, and delivered using the CloudFront CDN.
Because I had specific requirements on not using a
CNAME DNS entry to point
a CloudFront distribution, I also use Route53 to manage the DNS.
This costs me overall about 0.50 / 0.60 euros per month depending on the dollar-euro conversion rate, and taking into account that S3 and CloudFront cost me pretty much nothing since I am still eligible for the 1-year free tier.
That is a pretty good deal I guess, and from my traffic figures it should not be significantly more expensive when I exit the free tier.
You should also pay attention to the length of your SSL Certificate: it should be no more than 2048 bits. I initially had a 4096 bits certificate, and spend some time digging the issue.
Setup, uploading and invalidating
Uploading to S3 and invalidating CloudFront entries can be daunting, so you should simplify your task with s3_website.
It also does the proper configuration of your S3 bucket to server websites over HTTP. It does the same with CloudFront, by creating a distribution if need be.
It uses a simple
s3_website.yml configuration file. In my case it looks like this:
Note that I am using
PriceClass_100, which means that I chose the cheapest CloudFront price class.
Most of my traffic comes from the US and Europe anyway, so I can save money by not having broad CDN replicates
around the globe.
s3_website.yml is likely to be in the root of your Jekyll content, so it is a good idea to
exclude it from the output files in