Every data hoarder has the same origin story: one day something small disappeared — a project folder, a music collection, an old hard drive — and the loss stung enough that you swore never again. Now you have terabytes, spread across Google Drive, Dropbox, OneDrive and a couple of buckets, organized with a librarian's love.
And yet, if you're honest: most of it exists in exactly one cloud each. That's not a backup strategy. That's diversified risk of loss.
The rule, adapted for clouds
The classic 3-2-1 rule (3 copies, 2 media, 1 offsite) maps cleanly onto cloud life: every dataset that matters should live on two unrelated providers, with one automated pipe between them. "Unrelated" matters — two folders in the same Google account fail together. Google → Backblaze or Dropbox → S3 fail independently.
Why mirrors, and why they're scary
For an archive, you want mirror mode: destination is an exact replica — new files copied, changed files updated, deleted files removed. The deletion part is what makes mirrors trustworthy (no zombie files) and also what makes them dangerous: a bad selection or a moved folder means the mirror happily deletes on the other side too.
This is why the tool you pick needs two specific features:
- A dry-run preview. CloudSync shows exactly what a mirror run WOULD copy, update and — critically — delete, before anything executes. Run it after any big reorganization; it turns "I hope this is right" into a diff you can read.
- Failure alerts that reach you. A silent failure discovered eight months later is a lost archive with extra steps. CloudSync auto-retries transient failures three times, then alerts by email — or Telegram/Discord/webhook if, like most hoarders, you live in one of those.
A reference setup
- Primary → cold copy: Google Drive (working data) mirrors nightly to a Backblaze B2 bucket (~$6/TB/mo). Incremental comparison means unchanged files are skipped — nightly runs on 2 TB take minutes, not hours.
- Dropbox → second bucket or region: same pattern for your second cloud. Each dataset gets exactly one mirror rule; resist the urge to mirror mirrors.
- Filters keep the noise out: exclude
*.tmp, *.part, node_modules-style clutter per rule. - Weekly digest as heartbeat: CloudSync emails what moved this week — files, bytes, failures. If the digest ever goes quiet, something upstream broke. It's monitoring you don't have to build.
- API for the tinkerers: every rule can be triggered via REST with an API key — wire syncs into n8n, Home Assistant, cron on a Pi, whatever your lab looks like.
Total cost of actually sleeping well
2 TB mirrored to B2: ~$12/month of storage plus a Pro plan. Compare that with the going rate of the data itself: irreplaceable. Set it up once — the robots run nightly, the previews catch your mistakes, and the alerts find you when reality disagrees with the plan.
Start with one mirror rule on the free plan and watch the first dry-run preview. Hoarders tend to fall in love at that screen.