Expectations when migrating from self-managed to managed Postgres

Questions / Help
, ,
1

Hi, we’re planning to migrate a production app from self-managed Fly Postgres (postgres-flex) to Managed Postgres, Basic plan, PG 17, region sjc. Before we do, could you confirm:

  1. How often MPG takes automatic backups, and how long they’re retained?
  2. The length of the point-in-time recovery window?
  3. Expected failover time (RTO) and potential data loss (RPO) if the primary fails?
  4. Typical restore time for a database under 1 GB?
  5. Who applies minor Postgres security patches today, and whether there’s a maintenance window we’d be notified of?

Thanks!

2

Hi there,

  1. Backups: A full backup runs daily, a differential every 6 hours and an incremental every hour. Transaction logs are archived continuously. Backups are kept for 10 days.
  2. Point-in-time recovery: You can restore to any point in the last 10 days, up to about 1 minute ago. A restore always creates a new cluster.
  3. Failover: If the primary fails, a replica in the same region takes over automatically usually under 45 seconds. Replication is asynchronous, so you could lose the last moment of writes, usually under a second.
  4. Restore time for a database under 1 GB: Typically a few minutes, mostly spent provisioning the new cluster.

Cheers,

3

Thank you for the quick reply. Do you happen to know the answer to 5. who applies Postgres security patches and whether there’s a maintenance window we’d be notified of?

much appreciated!

4

Hi!

Sorry I missed that one. We apply Postgres patches, there’s a maintenance window you can configure (check settings for your cluster, by default it’s on Sundays). We do any needed operational work including updating Postgres periodically during that window only, you should plan for possible short downtime during that window, e.g. for lockstep restart of your primary and replica, pooler and other components.