Skip to content
Rowsafe
Docs

What Rowsafe does for you

What runs by itself, what you do on a bad day, and what Rowsafe never does.

Once your database is set up (Quickstart), most of Rowsafe runs by itself. On a bad day, you click and Rowsafe does the work.

Every day, by itself

  • Every change is saved to your bucket, at most about a minute behind, so you can go back to any second. Backups and recovery. OpenSearch, Qdrant and Meilisearch get a snapshot every 5 or 15 minutes instead.
  • Nightly backups at 01:00 UTC: full on Sundays, only the changes other nights. Two full backups kept (about two weeks). Old ones are removed for you.
  • Proof every Sunday evening: restores your latest backup into a scratch copy, checks it, deletes it.
  • Pulse every minute: disk, connections, slow queries, cleanup, backups. A health score 0-100, findings in plain words, and Apply fix for most.
  • Alerts when something needs you, by email, Slack, Discord, Teams, PagerDuty, Opsgenie, Telegram, ntfy or webhook. Add a channel under Settings → Alerts and notifications (channels).
  • Your weekly Pulse email every Monday.

Everything is encrypted on your server, with a passphrase only you have.

When something goes wrong

You deleted rows by mistake

Someone ran DELETE FROM customers with no WHERE. Production keeps running while you fix it:

  1. Rewind → Find the moment, then Rewind to just before this.
  2. Restore a copy as it was then, next to production.
  3. Compare with production to count missing rows per table.
  4. Bring back the missing rows. Newer rows stay as they are.

A bad UPDATE? Same steps, with Also restore rows that were changed.

An AI agent's migration broke the database

With Guard, the agent set a Mark right before the migration and told you its name.

  • On Marks, click Restore a copy at this Mark, then compare and bring back rows.
  • A dropped column or table can't be brought back as rows. Rewind the whole database to the Mark instead. What was written after is kept aside for 7 days, so you can undo.

Next time, preview the migration on a copy first.

The disk is filling up

Pulse warns up to 14 days ahead. If Rowsafe can clean up the cause (an inactive replication slot, unused indexes, tables that need cleanup), the finding has Apply fix. If your data simply grew, you need a bigger disk. Disk forecast.

Backups stop working

For example, your bucket keys changed. A critical alert fires within about 10 minutes and the database stops counting as protected. Fix the cause, then click Check again.

The server dies

  • With a standby: Standby → Promote. Automatic failover is available, off by default.
  • Without one: your backups are safe in your bucket. After 10 minutes of silence, click Restore to another server on the database: it restores onto another of your servers with the same storage and passphrase. Server lost.

What Rowsafe never does

  • Never sees your data. Backups are encrypted on your server; passphrase and keys stay there. Rowsafe stores only names, sizes, settings and metrics. Security model.
  • Never restarts or changes your database on its own. Only when an owner or admin (or their AI agent, as them) clicks and confirms, and only what the server's root allowed. Exceptions you turn on yourself: automatic minor updates and automatic failover. Rowsafe Cloud servers get security updates automatically, and restarts happen in a weekly maintenance window with an email the day before.
  • AI agents never get more than their person's rights. An owner's or admin's agent acts right away, with a Mark first. Otherwise Rowsafe refuses and says why. Every agent change is audited. How AI agents act.
  • Never runs commands or SQL sent over the network. The agent runs only a fixed list of tasks.
Edit on GitHub