SOP: Pausing Raxx dev compute

When we stop developing for the night / weekend / for a week, scale Heroku dynos to 0. This is the dominant cost lever. Everything else (Cloudflare Pages, DNS, Postgres essential-0) is free or $0.007/hour.

Toggle

# Pause everything paid
./scripts/ops/dev-pause.sh

# Resume staging only
./scripts/ops/dev-resume.sh

# Resume staging + prod (rare — only when deploying a release)
./scripts/ops/dev-resume.sh --with-prod

Requires HEROKU_API_KEY or HEROKU_API_KEY_PROD exported in your shell.

Service	State	Cost
`raxx-api-staging` web dyno	scaled to 0	$0
`raxx-api-prod` web dyno	scaled to 0	$0
`raxx-api-staging` Postgres (essential-0)	attached, idle	~$0.007/hr, $5/mo cap
`raxx-api-prod` Postgres (essential-0)	attached, idle	~$0.007/hr, $5/mo cap
Cloudflare Pages (`raxx.app`, docs, marketing)	live	free tier
Cloudflare DNS	live	free
`api.raxx.app` / `api-staging.raxx.app`	DNS still resolves, but returns 503 from Heroku router	free

Rough ceiling while paused: ~$10/month (Postgres × 2). Everything else is $0 until dynos come back.

What to NOT do to "save more"

Don't destroy Postgres add-ons unless you're pausing for months. Re-attaching rewrites DATABASE_URL, and you lose any data (currently empty; will matter once users sign up).
Don't remove the custom domains from Heroku. They don't cost anything idle, and re-adding them forces a new Let's Encrypt cert cycle.
Don't disable Cloudflare proxy on raxx.app. It's free and the SSL cert validation depends on it.

When someone hits the site while paused

https://raxx.app — Cloudflare Pages still serves the Antlers build (static HTML + JS). The app loads. API calls fail with 503.
https://api.raxx.app/... — 503 "App sleeping or no web dynos running" from Heroku router.
https://raxx.io — (future) docs site, always up.

If you want a friendly "paused for maintenance" banner instead of 503s, that's a console.raxx.app feature — toggle lives there. See the architecture doc once filed.

Morning checklist (resume dev)

./scripts/ops/dev-resume.sh
Wait for the script to confirm /api/system/status returns 200 (~30s cold start)
Open https://raxx.app and verify the dashboard loads (no red toasts)

When to add prod to the toggle

Default: prod stays at 0. Only scale prod up when you're actively cutting a release (v*.*.* tag pushed → GH Actions will bring it up anyway). If you need a hot prod for an incident or QA session, run:

./scripts/ops/dev-resume.sh --with-prod

Fallback: emergency kill (everything down including Postgres)

Only if we need to zero out all bills mid-month:

heroku addons:destroy heroku-postgresql --app raxx-api-staging --confirm raxx-api-staging
heroku addons:destroy heroku-postgresql --app raxx-api-prod     --confirm raxx-api-prod

This deletes the database. Use only when the data is either empty or has been exported. To restore, re-attach with:

heroku addons:create heroku-postgresql:essential-0 --app raxx-api-staging
heroku addons:create heroku-postgresql:essential-0 --app raxx-api-prod

New DATABASE_URL will be set automatically.

Raxx · internal docs