prod-deploy

name	prod-deploy
description	Deploy services to feedsubscription.com production. Use when the user asks to deploy, push to prod, restart a container on prod, or rebuild an image on production. Handles git push, SSH ControlMaster setup, and running make targets remotely.

Prod Deploy

Deploys one or more services to feedsubscription.com.

Standard workflow

1. Push commits

git push

2. Establish SSH ControlMaster

ssh -M -S ~/.ssh/control-feedsubscription -o ControlPersist=10m -fN feedsubscription.com
echo "ControlMaster established"

Skip if already established (the command will silently succeed or fail harmlessly).

3. Deploy

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'cd ~/src/rss-email-subscription && git pull && make <targets> start'

Replace <targets> with the services to rebuild before restarting.

Service → make target mapping

Service	Target	Notes
app	app	Also rebuilds delmon (FROM app)
certbot	certbot
logger	logger
website	website
smtp-in	smtp-in
smtp-out	smtp-out
resolver	resolver
delmon	delmon	Or just rebuild app

make start (always included) runs docker compose up --remove-orphans --detach and restarts only the containers whose images changed.

Examples

Deploy only app:

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'cd ~/src/rss-email-subscription && git pull && make app start'

Deploy certbot, logger, and website together:

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'cd ~/src/rss-email-subscription && git pull && make certbot logger website start'

Deploy all images (full rebuild):

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'cd ~/src/rss-email-subscription && git pull && make all-images start'

Smoke test (optional)

If the user asks for a smoke test after deploying app, trigger a feed check by removing lastPostMetadata.json and sending HUP:

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'rm src/rss-email-subscription/.tmp/docker-data/accounts/*/feeds/gurdiga/lastPostMetadata.json && docker kill -s HUP app'

Then confirm the sending report appeared in app.log (wait ~15s first):

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'grep "Sending report" src/rss-email-subscription/.tmp/logs/feedsubscription/app.log | grep "gurdiga" | tail -3'

Look for "sent":N,"failed":0 in the report data.

Then confirm delivery via smtp-out log:

ssh -S ~/.ssh/control-feedsubscription feedsubscription.com \
  'docker logs smtp-out 2>&1 | tail -40'

Look for status=sent (250 2.0.0 OK) for each queued message, and removed confirming the queue cleared. Any status=deferred or status=bounced lines indicate a delivery problem.

Notes

• Always git push before deploying so git pull on prod gets the latest commits.
• make start is always appended — it's a no-op for containers whose image didn't change.
• Timeout: allow up to 5 minutes for builds that compile TypeScript (app) or download pip packages (certbot).
• After certbot rebuilds: verify with docker exec certbot python -c "import certbot; import OpenSSL; print('ok')".