- Published on
The rabbit hole called docker networking
- Maik Diepenbroek
I run a native version of nginx on my server, on it i have docker installed, and a mix of manually set up letsencrypt certificates and virtual hosts.
For running this website, i thought of modernizing that a bit by dockerizing it. No sooner said than done i created a docker version of my website, created a proxy rule in nginx and with high hopes i reloaded the nginx configuration by running
nginx -s reload.
After looking at a spinner, and performing some hard refreshes i still got nothing but timeouts and eventually an error page.
So i started digging in, my approach is written down in the next paragraph(s)
Path to the solution
- Started with the logs, which unfortunately told me nothing
- Changed the proxy pass rule to use the docker0 bridge IP, also no luck there
- See that the application is running on localhost, so i manually altered the
/etc/hostsfile to do some trickery, also (as you might've imagine) no luck
- Reading up on what the experts have to say about it, but without a real knowledge on what to look for, that only confused me more. Which has nothing to do with the excellent articles by the way!
- Dockerize my complete nginx setup, which worked for a single site but since my other applications weren't dockerizable that resulted in nothing but respect for the technologies used.
- Diving into host modes, bridges, macvlan and NATs which goes way over my head so i felt like a fish out of water.
- Running the application on
localhostalso didnt do the trick which made me understand the underlying issue even less.
Cowards way out
I ended up using the
--network=host option to utilize my own network stack which did solve the issue but i'm still not happy with.
Any suggestions are more than welcome.
I could not let it go, so took one more stab at it (earlier this year). I decided to simply create a custom bridge network, with this all my issues went away and i could continue to use docker the way i was used to.