WASSUP HOMIES

How to get it all going

Run it locally

just runme # https://just.systems/

prod

I did these on Ubuntu "24.04 LTS (Noble Numbat)"

As root:

yes '' | adduser --disabled-password --quiet teensy
DEBIAN_FRONTEND=noninteractive apt install -y git nginx snapd python3-poetry

• su - teensy

mkdir ~/shorty
cd ~/shorty
git clone https://gitlab.com/offby1/teensy.git .
mkdir -vp ~/.config/info.teensy.teensy-django
cat > ~/.config/info.teensy.teensy-django/.env

• Now paste the env file:

  • grab the RECAPTCHA_SECRET from https://www.google.com/recaptcha/admin#site/320420908, log in as me, click the gear, click the "reCAPTCHA keys" thing, click "copy secret key"
  • generate SECRET_KEY with python3 -c 'import secrets; print(secrets.token_urlsafe(100))' In a perfect world, if you're moving the site from one host to another, you'd use the same SECRET_KEY on both, since I think that means that auth tokens would then transfer over. But on the other hand, the only person who needs to authenticate is me, so ... 🤷

• exit # back to root

• cat > /etc/systemd/system/teensy.service

Then paste the file of that name from this directory

snap install lego
lego --email="[email protected]" --domains="teensy.com" --http run

• scrape generated cert and key outta wherever they wound up -- /var/snap/lego/common/ iirc -- , and plop 'em in /etc/pki/nginx, where the config looks for it
• yeah this should probably all be a recipe (or recipes) in the justfile

TODO

• Set up cron job to run sync-ddb-data, as above. DJANGO_SETTINGS_MODULE=project.prod_settings nice ~/git-repos/me/teensy-django/.venv/bin/python manage.py sync-ddb-data will probably do it.

TLS

• Consider https://github.com/NginxProxyManager/nginx-proxy-manager instead of lego &c
• Similarly https://caddyserver.com/
• https://github.com/amerkurev/django-docker-template

Easy Auth

• https://www.photondesigner.com/articles/email-sign-in

DONE

• figure out how to deal with Docker secrets

  https://docs.docker.com/engine/swarm/secrets/#build-support-for-docker-secrets-into-your-images suggests it's straightforward—and in fact I've done something similar in the docker-take-three branch. It's confusing, though, since much of the documentation around docker secrets refers to "docker swarm", which is ... I dunno? An alternative to docker-compose?

• Go through the nginx beginner's guide slowly and carefully Not very informative, as it happens :-|

• plop it on an actual EC2 box and test it in "production" mode. In particular: the recaptcha

• Come up with a better django admin password, and a better mechanism for getting it onto the prod host The password is now in bitwarden and Firefox. The "better mechanism" is just running DJANGO_SETTINGS_MODULE=project.prod_settings poetry run python manage.py changepassword ubuntu at the command line.

• Tell Google not to index the site

• Tweak gunicorn logging so it puts the actual IP address in the log, not 127.0.0.1

• Again look into replacing runme.sh with "just"

• Consider whitenoise instead of a special section for nginx

• Test with rudybot! Pretty sure I need to whitelist its IP address. A recent log against the pyramid server looks like 144.217.82.212 - - [14/Jul/2024:19:09:16 +0000] "GET /shorten-/?input_url=https%3A%2F%2Fmy.what.a.long.url%2Fyou%2Fhave%2Fgrandma%2Fmy.what.a.long.url%2Fyou%2Fhave%2Fgrandma%2Fmy.what.a.long.url%2Fyou%2Fhave%2Fgrandma%2Fmy.what.a.long.url%2Fyou%2Fhave%2Fgrandma%2F HTTP/1.1" 200 30 "-" "Racket/7.9 (net/http-client)" fwiw

• Update the nginx.conf again, to have just one server

• Figure out why my old "teensy-2022" host died :-| tl;dr: StartLimitAction=reboot 🤢 All I remember:

  • I was fiddling the teensy.service file, and did something like sudo systemctl start teensy to start it
  • systemctl said something like "golly I noticed some config files have changed; please do systemctl daemon-reload the world or something
  • from that point on it was weirdly unresponsive -- CPU usage went to about 60%, and I couldn't ssh in
  • perhaps attach that old root disk to the new host, and poke around in the logs

• Update the systemd file.

Rewrite notes

Rewrote it in Django -- it used to use Pyramid.

Uses sqlite instead of dynamodb:

• django doesn't work well with dynamodb (or any nosql database);

• sqlite otta be fine given how little data we have

  As of 2024-07-01T08:09:41-0700:

  Item count
  10,689
  Table size
  2.2 megabytes
  Average item size
  204.17 bytes
  

  We will back up sqlite by having "cron" or "systemd" or whatever run python manage.py sync-ddb-data every now and then. Or, you know, I'll just run that when I think of it :-)

  I have another management command, backup-db-to-s3, that isn't needed any more, but I feel like keeping it around for some reason.