You don't need a $500/month observability bill when you're a one-person operation. You need boring tools wired together correctly.
Here's my complete stack — every piece free, total monthly cost $0, and it has caught every revenue-affecting failure for the last several months.
Layer 1: Process supervision (free, built-in)
launchd on macOS / systemd on Linux. Both restart dead processes automatically and log to journald. This isn't a 'tool choice', it's the floor: anything that must run 24/7 runs under a supervisor. If you're using nohup for anything critical, stop today.
Layer 2: Health sentinel (Python + JSONL, 40 lines)
A single loop: hit every critical URL every 5 minutes, append status+latency to a JSONL file, alert only on state change. JSONL over a database because grep is a perfectly good query language for one human, and the file never needs migrations.
Layer 3: The dead-man's-switch (free tier of any uptime service)
Your monitor monitoring itself: POST a heartbeat every minute to an external watcher. 3 missed heartbeats = the watcher emails you. This is the layer people skip, and it's the one that catches the day your whole box silently dies.
Layer 4: Self-healing loop with an attempt budget
On failure: restart process, clear cache, re-auth session — max 3 attempts, everything logged with evidence. Only then does a human get pinged. Roughly 90% of incidents die at attempt 1 and never reach me.
Layer 5: The morning digest
One message per day: checks run, failures, auto-heals, needs-human. Live alerts only for fires; the digest covers everything else. Interruption load dropped ~90% after I moved to this split.
Why JSONL + grep beats a dashboard at this size
Dashboards rot. Grep never does. When every external signal — sale, click, payout, failure — lands in one append-only log with a channel tag, you can answer 'which channel produced this money?' in one command. Attribution at $0/month.
If you want my exact field cards, runbooks, and copy-paste code for all 5 layers, I sell them as kits on Gumroad — the Automation Starter Pack is the entry point:
The full collection (ops kits, runbooks, monitoring bundles) is in the Ops Mega Bundle:
Happy to answer questions in the comments — and if you have a $0 stack of your own, share it.
Top comments (0)