managed postgres

Questions / Help
1

I’ve been looking at managed postgres providers. One that looks really great is Aiven and they have hosted postgres on a few different cloud providers (not fly tho). I was wondering if there’s any reason to pick aws vs azure vs gcp vs digitalocean as the host for postgres. Maybe there’s better latency from fly to some cloud providers than others? Better connectivity? Maybe there’s other postgres hosts that do work with fly? I’m looking into a fully managed postgres service, rather than running postgres myself on fly.

1 Like
2

Not sure what you mean by “managed”.

Postgres on Fly is the equivalent of another provider’s managed Postgres. With Fly you literally run one command, and it spins up a Postgres cluster, and that’s it, you can connect your app and/or the psql CLI and run queries.

Other providers that are managed also do this, there’s really nothing more a managed Postgres host will give you unless you mean like automatic vaccuming etc?

3

Hmm I sort of got the sense from posts like Future of Postgresql on Fly? that I should look into a more managed postgres service than what fly currently provides.

4

The distinction between “managed” and “automated” (what we call Fly Postgres) is a little fuzzy, but we’re erring on the side of under promising.

Managed usually means someone will get an alert when a service breaks, then drag themselves out of bed and fix the issue. We do manage the infrastructure underneath, but we do not respond to postgres specific issues (like running out of disk space or out of memory errors).

Automated postgres is enough for many apps! It’s very cheap to run, and works well for 99% of users. But some people think it’s worth paying 2-3x the price to have a on call DBAs jump in and fix things if they break. I don’t fault these people, it makes a lot of sense to spend money on that kind of peace of mind.

Aiven does seem great, so does Crunchy Bridge. If you want that level of service, you can happily use these with your Fly apps. For the most part, they’ll be <1ms round trip latency from cities we share with AWS/GCP. Fly.io IAD and AWS us-east-1, for example, work just fine together.

1 Like
5

As one of the co-founders of mongohq/compose, I’m sure you have more insight into this… But to be honest, I’d rather not manage any database, automated or self-hosted, for prod apps. It simply isn’t worth it.

1 Like
6

Hi @kurt, what data centres does Fly share with AWS? Looking to setup AWS RDS (or any other managed DB provider really) in the same data center as my Fly apps for that minimal latency. Looking for one in Europe, US and Australia.

7

hi @seva,

Fly.io does not share datacenters with AWS. What I would recommend is looking at our list/map of regions and picking the ones closest to your RDS or managed DB datacenter locations. For example, for Amazon’s us-east-1 the closest Fly.io regions might be iad and atl.

-Daniel

2 Likes
8

For anyone looking at this in 2025:

Fly.io is building a Managed Postgres product called Fly MPG, which should have a first version around Feb or Mar 2025:

4 Likes
9

As a small bit of news, this has an official page now—including the first information on pricing:

https://fly.io/docs/mpg/overview/

(Minimum of ~$40/month, if I’m reading the table correctly.)

3 Likes
10

When is managed Postgres expected to launch in the NRT region?

11

We don’t have plans to expand Managed Postgres regions just yet.

12

I’m trying to create a managed postgres cluster using

fly mpg create --volume-size 15

But it still creates a 10GB cluster. The issue is pgloader bails out with out of disk space error:

ERROR Database error 53100: could not extend file “base/16466/16540.6” with FileFallocate(): No space left on device
HINT: Check free disk space.

13

Hello Fly Team!

Any thoughts about a Managed Postgres plan that falls between ~$40 and ~$300 / month? That’s a huge jump, and I’m guessing that I’m not the only one that might need a bit more than a 1GB Basic cluster but not nearly as much as an 8GB Launch cluster.

I’m currently running a 3-node Legacy Postgres cluster with shared CPUs and 4GB of memory and it’s working perfectly (though I’d like to move to Managed since I just spent the morning dealing with a full disk that caused chaos and I’d rather not have that happen again).

14

This should be fixed in flyctl v0.3.145. Thanks for the heads up!

15

Heya - thanks for the feedback! We’re currently most focused on nailing some basics so I don’t really have a timeline or commitment for you. I’ll chat with the team, but we’re definitely sympathetic to wanting some more options that better fit that kind of need.

2 Likes