Skip to content

Latest commit

 

History

History
83 lines (53 loc) · 2.94 KB

FAQ.md

File metadata and controls

83 lines (53 loc) · 2.94 KB

FAQ

If you have a question create an Issue or write me somewhere.

How did I setup wireguard?

Using this project: k4yt3x/wg-meshconf.

It manages a CSV containing all involved nodes and can generate the config and keys for all nodes. I then use ansible to move them into place and start the tunnels.

Why Wireguard instead of VPC peering/transit gateways?

In my professional life I deal a lot with these VendorOps solutions because most companies want them but in my private life I try to learn the technology/protocols behind them.

Initially I planned to run each VM on a different hoster somewhere in the world but finding them, dealing/paying them and making sure they support FreeBSD was taking too much time, so I went back to AWS.

Why FreeBSD?

Let me start by saying: This is my opinion. You can have a different one. That's fine. :)

Linux distributions took away the fun with Linux for me. Every new release meant re-learning a lot of new things and breaking changes left and right.

Two years ago a friend recommended FreeBSD as an alternative. For a year now I'm using it primarily on my servers. It's a very simple OS. It's capable, fast and does what I need. The documentation/handbook is the best thing I've seen in IT in the last years. I miss docker from time to time though.

But as with Linux, I wouldn't dare to run it on my notebook as a daily driver.

Keep in mind: This is my opinion. YMMV! :)

Why packer?

The default AMIs for FreeBSD do a system upgrade on the first boot which takes somewhere between five to ten minutes. This is too long to stay sane, so I created a custom AMI that has this step already done.

Some FreeBSD people are currently making this situation better.

Why not fly.io/gigaelixir?

I had a couple of things running on fly.io in the past and it was ok, but ... I want to learn stuff that I can use whereever I go.

All these abstractions come at a cost. Once something doesn't work you have to deal with a company. For me this feels like wasted time. Again, YMMV.

Why no TLS on the site?

I'm lazy. The infrastructure work already took way more time than I'd like to admit. Remember, it's a side project that will go away soon.

What else is involved?

Each VM also runs HAProxy and a Grafana Agent.

Grafana is configured to use Grafana Cloud, but the metrics are rather useless because there is barely any load on the VMs.

How much does it cost to run this on AWS?

About 50 Euros per month.

Oh my gosh those numbers. Are you famous?

No. Don't trust these numbers. I regularly ran load tests using wrk.

I did this because my first attempts with this app resulted in a very slow web site that could handle around 1.25 requests per second. You can read about it in the changelog.

I also wanted to see how traffic between the servers flows when one of them is hit by a lot of traffic. Interesting stuff.