Managed Postgres in sin: HA on Basic, SLA, backup retention, and origin auth from Cloudflare

Questions / Help
1

Hi! I’m evaluating Managed Postgres in the sin region for a small production
app (Django + Next.js, currently pre-launch). Four things I couldn’t settle
from the docs:

  1. HA on the Basic plan
    The MPG docs say all plans include high availability, but Basic is described
    as a single shared-2x node. Does Basic include automatic failover to a
    standby, or does “high availability” mean something narrower there? If
    failover does exist on Basic, what’s the expected RTO, and is the standby on
    a separate host/zone within the same region?

  2. Product status and SLA
    Is MPG generally available, or still beta/preview? I couldn’t find a GA
    announcement. Is there a published uptime SLA? If not, what availability do
    you target internally?

  3. Backup retention and PITR
    What’s the default backup retention per plan, and the maximum
    point-in-time-restore window? Is retention configurable? Are backups stored
    in the same region as the cluster?

  4. Origin authentication from Cloudflare
    We plan to put Cloudflare in front of the app. Is there a supported way for
    a Fly app to verify a request actually came from our Cloudflare zone — e.g.
    client certificate verification at the Fly proxy (Cloudflare Authenticated
    Origin Pulls) — or is that meant to be handled inside the app? Can a Fly app
    be made unreachable from the public internet except through Cloudflare?

Thanks!

4

Hello @MinLee! It’s excellent to read about your interest with MPG, thanks for all these great questions.

Let me answer each one below:

  1. All MPG plans include an automatic failover to a standby replica in the same region, which is deployed on a different host from the primary’s. Direct connections should still route to the proper primary, and a failover should be pretty quick.
  2. MPG is in GA, in fact, we recently moved our MPGv2 to General Availability too( see announcement ). While we don’t currently have an official SLA for MPG uptime–our team is dedicated to keeping the uptime as high as possible, we’ve got tooling and monitoring built to help with this. You can also escalate any issues from your dashboard’s MPG support portal.
  3. Backups are retained for existing clusters up to 10 days–these are stored in the AWS S3 region nearest to where the cluster is deployed in. While we have PITR for MPGv2 clusters, we unfortunately don’t have any true PITR for MPGv1 clusters. Of course, you can grab any of the existing backups and restore unto a new cluster.
  4. We don’t have an out of the box Fly Proxy feature that handles Cloudflare authenticated origin pulls. This would have to be handled from within the application you deploy. It’s also possible to remove a Fly app’s Public IP address–and reachability from the internet–in fact we have a whole section for deploying private apps here. And while there’s no in built feature to allowlist certain IPs or hostnames, perhaps you can look into setting up a Cloudflare tunnel!