How Munchable Built a Smarter Postgres Backup Job That Fails With Clear Errors

Munchable, whose Postgres database lacks managed backups, relies entirely on a nightly GitHub Actions job to dump data to Cloudflare R2 and alert admins on failure. The backup is split into two archives: a full dump retained for 30 days and a smaller core dump — excluding the rebuildable product catalogue — kept for a year at near-zero cost within R2's free tier. After the first run failed with an uninformative 'Network is unreachable' error, developers added preflight checks that detect and explain common misconfigurations, such as using an IPv6-only direct host or the transaction pooler port that pg_dump cannot use. Each error message names the specific fix and points to a runbook file, so on-call engineers get actionable guidance rather than a bare failure. Bucket lifecycle rules, not the workflow itself, handle backup deletion to avoid accidental data loss from scripting bugs.