Dockerise and Kamal 2

[leonardow-unep-wcmc]

Changes

• Ruby: Upgraded from 2.3.1 to 2.6.10.
• deploy/: Added a folder containing its own Gemfile and .ruby-version to pin the Ruby and Kamal versions for deployment.
• Kamal configs: Added Kamal configuration files.
• Kamal pre-build hook: Enabled the pre-build hook in Kamal.

Notes

• In the Dockerfile, the Ruby image must specify --platform=linux/amd64 because the libv8 gem does not support ARM.
• Run docker compose up -d to start the local development environment.
• The final Docker image for both staging and production is based on Nginx.

Nginx

• Although this repo README.md mentions an Nginx rewrite, I can’t find it on the staging or production servers, so I assume it isn’t necessary.
• The official Nginx default configuration is used, which should allow serving any domain on port 80.
• In the current staging/production Nginx configuration, Passenger is running with Ruby 2.2.2 and has a special handler for the /assets/ folder. However, based on my testing, this is a JavaScript-only frontend project—it doesn’t require Ruby at runtime and there is no /assets/ directory. It’s likely that a previous DevOps engineer simply reused a configuration from another project. That’s why I insist that we include exactly what’s needed—no more and no less—rather than piling on every setting that “seems” harmless. Future developers or DevOps engineers will otherwise have a hard time determining which directives are actually in use and which can be safely removed.

TODO

• Remove Capistrano (not done yet).
• Finalise staging and production server addresses (TBC).

[pdl]

I have had issues - which Leo cannot reproduce - where some of js/app.js, js/minispade.js, js/handlebars.js randomly took > 30s and timed out. The workaround seems to be docker compose down -v (the resulting up will take quite a while as gems will need to be donwloaded).

[pdl]

Looks good to me, but need new machines to verify.