Here’s a quick #fresh-lettuce
One of the most common feedback we got in our newsletters about our Managed Postgres offering was that the increase in price from Basic ($38) to Launch ($282) was huge.
We semi-recently shipped a new plan called Started designed to be the middle ground between those running Shared CPUs x2 for each VM at 2GB of RAM priced at $72.
This should be enough to address some clusters that need just a tad bit more to work efficiently.
Let us know your thoughts about it here!
Any plans to add a Broke tier that’s half of the Basic?
We do want to add one like “development”-tier at some point but right now we are focusing on strengthening our current offering for production grade ones so I cant speak for an ETA.
Speaking plainly we are kinda concerned about abuse for lower prices to be honest. Having a fair and kinda-expensive price for the lowest offering helps us not have to worry about CC fraud and other kinds of abuse that would drain our team.
Loving Fly MPG so far!
What happened to the interface that lets you just … look at the database and query it? I swear it was there last week 
We discovered that there were some security concerns with the Explore tab, so we’ve disabled it for now. It’ll come back soon, we promise!
Ha ha, that made me giggle. I’d be in favour, as I have a long-running “I’ll make some money from this eventually” project, and I don’t want it to be a financial drain in the meantime.
But for now I am using the free tier of Supabase, and I have to say, it is pretty good. I think I get 0.5GB of storage, and for my minimal requirements, performance and uptime seem fine. I might run into latency problems in the future, based on it not being in the same DC, but for now, I recommend it.
500MB sounds like a decent size but that gets eaten up pretty quick if a workflow creates a lot of history. Or that might be a NEON issue? I’m not 100% sure.
My table is like 16MB but somehow it takes 800MB of total storage.
I assume the 800MB in your example is the way a database grows on disk optimistically and infrequently; it basically resizes for future expansion, rather than because it needs that size now. I’d hope the managed database providers are rather more conservative in their counting when setting usage limits 
vacuum it!
I tried vacuuming and everything else but it never worked unfortunately.
Is this cost per machine or for the cluster?
Also do you support Postgres 17?
Thanks!!
Cost is for the whole cluster, minus storage.
No PG17 yet, though we are starting to hear a lot of requests for it. Care to share the specific features you’re after?
I set up a single DigitalOcean Postgres with 1GB of RAM & 10GB of storage to share between all of my side projects. It’s $15 per month and includes backups & whatever else I would expect from a fully managed service. Instructions here: Connecting to a DigitalOcean Managed Postgres from your Elixir Phoenix app on Fly.io .
for a better comparison you need to add an extra node to DO postgres, prices are almost same if you do that. Fly also gives you connection pooler per database and makes it very easy to create logical dbs under one cluster
DigitalOcean also does not charge for connection pools, and I have created one per logical DB, at no extra cost.
If I would indeed get a 2-node Postgres cluster from Fly.io for $38, then I would need to update my pricing comparison to be fairer and just provide it as a hack for low-budget side projects.
I’m looking at these pages:
And neither of them clarifies that all of the pricing is for a 2-node cluster.
yeah fly mpg is HA meaining its 2 nodes by default even the cheapest one, you are charged per use so you can spin one to try. I gave it a go and they did great tbh
That’s great to know, and should probably be made more clear to show your value proposition against competitors.
Also, if scaling down to 1 node and being charged $19 per month is possible, it should at least be documented as an option, probably with the caveat: “Only do this for hobby projects, you might have a short outage during upgrades. But your data will still be safe and you will still get all of the backup management etc.”
If I knew that I could provision a $38 Postgres cluster and then scale down to 1 and be charged $19, I probably would never have spent the time on figuring out how to connect to DigitalOcean.
Scaling down to one node is not an option for MPG clusters. All MPG plans are configured in a two-node, HA set up with automatic failover to the standby node in case of an issue with the primary. Each cluster also includes a 2-node PGBouncer set up.
We mention that the clusters are “High availability with automatic failover” in the docs, but sounds like we can make it clearer what that entails. Thank you for the feedback!
I updated the linked blog post to be more fair to Fly.io, now that I understand I was comparing a 2-node HA cluster vs a single node.
