Is FlyIO the new Heroku for indie founders?
The month is November 2022. Heroku announces they'll be ending their free Heroku tier.
For indie hackers like us, that meant bad news. Very bad news.
Heroku was the go-to platform to test MVPs: Have a product idea? Spin a free Heroku tier in a few minutes and get things up-and-running. But now that workflow is… gone.
After a (bad) change, indie hackers do what they do best: They look for solutions. And one of the best ways to look for solutions is to start a discussion on Indie Hackers. That's what many did.
Two indie-friendly platforms quickly emerged: Render and Fly.io.
Here I'll review Fly.io (and why I found it more suitable than other services like Render for my needs as an indie founder).
My use case
I'm a Rails developer and do (mostly internal) micro-apps. I want an easy-to-use service that can run Ruby on Rails + some SQLite (for smaller apps) and PostgreSQL (for larger ones).
I also want the ability to background workers + have good uptime.
And most importantly, I want something where the costs are low. I have lots of apps that don't bring me any revenue, so paying max $5-10 per app/month is an ideal scenario for me.
Heroku simply didn't make the cut after they discounted their free tier.
Why I chose Fly.io (over Render and other PaaS)
First of all, the pricing pages. Here's the pricing page for Render. Here's the pricing page for Fly.io.
I don't know about you, but Fly's pricing page felt more intuitive to me like an indie founder. It felt more transparent.
It turned out that my intuition was right. I had to dig deeper to find that, for example, Render imposes a 15-minute inactivity limit for free VMs. Fly.io doesn't do this.
The free tier: With Fly I can theoretically run 3 apps for free and slowly add more resources as I go along.
I started doing this by the way. After a month, I've received this email from Fly:
That's a very nice touch from them.
Overall, I have a feeling that Fly.io values indie founders more than Render, even though they raised Series C (Render just raised Series B). It seems like those 2 companies took quite different paths.
Fly.io supports SQLite: Yup, I don't have to pay for a separate database VM if I run really small apps. That's a huge reduction in costs if you're dabbling with multiple small apps.
Closer to Heroku: This is Redern's page for deploying a Rails app. This is Fly.io's page for deploying a Rails app.
I've tried both and Fly felt closer to Heroku in terms of ease-of-deployment. I just had to run 2 commands: "fly launch" and "fly deploy" after getting familiar with the whole process.
It's (supposedly) faster: This is a Medium post where they compared Fly to Render and found Fly to be significantly faster. This could have changed by now, though.
It's not just me: I've been seeing some anecdotal evidence on Hacker News for founders preferring Fly.io over Render. It seems like Render wants to focus more towards enterprise-y customers with their overall offer.
Is Heroku still a viable alternative for indie founders?
Not if you're on a budget.
A quick look at their pricing page should get you the idea why.
"Production" VMs start from $25 and up. Decent Postgres VMs cost ~$50. In comparison, you can run fully managed Postgres instances on Fly.io for as little as $3.
So to summarize, Heroku could be a viable alternative for funded startups. But not for indie founders on a shoestring budget.
The downside of Fly.io: Downtime
I've been seeing HackerNews threads on Fly.io downtime for their VM servers and Postgres databases.
I've also experienced the same thing for my VMs. The status page was showing that everything is green but my site was unreachable for an hour or two.
This doesn't really make a difference since I have mostly hobby projects, but it will make a difference if you have a site where you need 99.99% downtime.
So this is one thing to have in mind after you deploy & grow and want to have a more "stable" platform.
The upside: They're not going anywhere
Crunchbase says that Fly.io received over $100M in funding so far and have investors like Andreessen Horowitz. They also have less than 50 employees, which is a sign they're keeping things lean.
I think it's safe to assume that Fly.io is not going anywhere in the near future. Any downtime issues they (currently) have will eventually get resolved.
Heroku was awesome and their free tier was awesom(er). Currently, Fly.io (despite its downtime issues) is the closest thing we have to it.
What's been your experience with Fly.io so far?
After 19,314 lines of code, i'm shutting down my project
Need feedback for my product.
Don't be a Jerk. Use this Tip Calculator.
40 open-source gems to replace your SaaS subscriptions 🔥 🚀
We are live on Product Hunt
The Heroku alternative for indie hackers is https://Appliku.com + Hetzner Cloud.
I host 18 apps on one $24/mo Hetzner ARM box for over a year now and it all works flawlessly.
I get all the fancy stuff like nice dashboard, super fast build times, databases, backups, crons, custom domains and everything is customizable too.
Even better - moving apps to a new provider in case of problems is a few clicks away and there is no need for drama about picking a new provider: just add a server from new provider and start placing your apps there.
With every fully managed PaaS it is always the same song - it is always way more expensive compared to any VPS and there are always "buts". By using Appliku you remove those "buts", because you have the ability to customize everything is you chose to or be happy with the defaults.
And yes, I am the maker of Appliku. I have dedicated already 4.5 years of my life to it, because I wasn't happy with any fully managed PaaS offerings. Still don't, by the way.
Happy deploying!
+1 to Appliku. I have been using it for my SaaS for over 3 years so far.
It is more stable, cheaper, and better supported than Heroku.
Very cool. Reminds me of Laravel Forge but for Python and JS apps instead of PHP.
Pretty much the concept, yes.
Basically the existence of forge and Hatchbox was enough validation that i needed.
you smelled it right🤣
I use Piku and I have 31 apps getting 1,000s of visitors per day without breaking a sweat for $12/m. Appliku looks super cool!
Yeah, Heard about Piku. Good job with the setup.
Basically there are options. Going from DIY on one side of the spectrum to fully managed on another side of the spectrum. Problem is fully managed makes you locked in with risk of them being unable to customise enough, to break down and you don't have any control over them, drastically raise prices etc.
As far as I remember DO Apps used to charge for the build time. They had to stop because they couldn't make it predictable and were timeouting all the time.
Yep it's definitely tricky to find the right balance, and the right balance is going to be different for different people too.
This is absolutely true.
Some need to be mindful about their costs, some need to make sure investors/other stakeholders are happy and would go to "proven" and managed solutions so they can blame someone when everything falls apart 🤣
For some being in the balance automation/convenience vs ability to customize is important.
Btw how big of a footprint is Piku on a server?
Piku is a single Python script with just under 2k lines of code. It doesn't run it's own background server. Instead it intercepts
git pushand relies on existing servers like nginx and uwsgi, so CPU costs are negligible. It doesn't consume any bandwidth and memory requirements are also very low, and only kick in when you push. Basically it is about as efficient as you can get and still call yourself a PaaS.Somewhere else somebody said they use a single bash script to deploy. I've read blog posts about this and I used to do it that way and it's a PITA. Piku is basically the same idea as "single bash script deploy" but with the shared knowledge of multiple people maintaining the open source project.
(Render founder) It’s true Fly’s free tier is more generous; OTOH, it’s also true that Render gives you much more beyond raw compute: native GitHub/GitLab integrations, truly fully-managed Postgres and Redis, static sites, cron jobs, PR previews, Slack integration, and more.
Ultimately, we want to empower you to move fast and maintain impeccable uptime. I was an indie hacker before starting Render, and helping individual developers avoid cloud complexity is a key part of our product DNA. If you’re happy and productive on Fly, that’s great. If you’d like to build on Render, we’re here for you.
I migrated to Render last year and don't regret it. We had major news go out last week that was picked up globally and drove a lot of eyeballs to the platform. Render didn't even bat an eye.
The support I've had over there has been awesome too.
(I do wish they'd support Laravel apps without the need for docker though @anuragg :-) )
So happy to hear that! Congratulations! This was a long time coming and I'm very excited for Songbox's future.
Native PHP support remains one of those things we want to do and will once we have a slightly bigger team!
Indie Hackers is hosted on Render, so if you're reading this, you're using Render :)
Fly gets a lot of buzz (patting myself on the back for that sweet pun), but I've seen a lot on HN about it being unreliable, going down, etc. over the years. In fact I just searched HN and this is from 8 days ago: Fly.io Postgres cluster down for 3 days, no word from them about it .
I also like that Render has a beautiful UI, is focused more around using the web UI than a CLI, and auto-deploys when I push to my Github repo.
I moved from AWS last year.
I had a terrible experience on fly, with both the service and the support.
Then I moved to Render and it's been great. Couple minor hiccups but nothing much.
I've heard very positive things about the developer experience for both Fly.io and the Render, but have heard bad things about the stability of Fly.io and nothing but positive things about Render.
It's worth recognizing that the Fly.io team's main "bet" isn't around "creating the best possible developer experience", but rather the idea that hosting is going to be pushed to the edge in order to get reduced latency for calls.
I'm actually pretty bullish on this idea, but it comes with a terrific amount of technical complexity and the Fly.io team has had a lot of problems keeping that "bet" stable with an influx of users.
Render, on the other hand, is a closer spiritual successor to Heroku's "backend hosted in a single location" which is a less ambitious idea technically but also probably easier to scale up with new users.
If it were me, I think I'd probably choose Render, at least until Fly.io shows that they can handle the influx of users without significant stability problems.
I tried both but had to upgrade to paid on Fly for the app while I could make it work for free on Render. Neither is really better at this point.
I just use $5 VMs (one per project) and little bit of Bash. That's it.
I don't think Heroku is overpriced for indie hackers. You can get very far with the starter plans (for example Basic 7USD + Postgres Mini 5USD), and reliability is very good. If you need the higher plans with a lot of instances, it's getting very expensive indeed.
Yeah, I'm a bit confused by this as well. If Heroku's free tier is $7 and that is considered too expensive, it makes me skeptical about whether a business offering less than that can be sustainable, especially if they're charging pennies per customer.
I can understand why you feel Heroku might not be overpriced for indie hackers and small teams. The starter tiers offer decent capabilities at reasonable costs, especially when reliability and ease of use are top priorities. As projects and traffic grow, the costs scale up rapidly on Heroku. But for early stage apps and side projects, the free and low-cost plans provide good value.
The key advantages of Heroku like auto-scaling, add-ons ecosystem, and simple deployments can outweigh the premium pricing. For larger scale production apps with big resource needs, Heroku does get very expensive compared to alternatives. But for smaller scale apps where developer productivity and convenience matter more than raw infrastructure costs, Heroku remains competitive. The pricing tiers give flexibility to scale up smoothly as needed.
So I agree Heroku merits consideration for indie developers and startups, especially in the early and growth stages. The premium for convenience and reliability may be worth it before projects reach big scales. But once resource demands increase substantially, it makes sense to reevaluate the cost efficiency of alternatives like AWS, GCP etc. Heroku fills an important niche, but isn't necessarily the most cost efficient at large scale.
Hey! I also tested both platforms a few weeks ago. I chose Render for my “stable project”, mainly because it was easier to setup a cron job there than in Fly. I wasn’t aware of that downtime issue, to be honest.
For another project I really needed a specific version of libvips installed so the fact that Fly uses a Dockerfile for the deployment made everything easier for me. I also love those emails in the end of the month :)
BTW, I’m a Rails developer as well.
You can use Dockerfiles with Render: https://render.com/docs/docker
Yep, you can now. I believe it wasn't an option a few months ago.
As others have said I think Render is closer in spirit to Heroku, I'm using Fly specifically since I have some latency requirements, and they make edge deployments and clustering in Phoenix very easy.
I've never heard of this before, but it sounds like something that would work for me. I'll check it out, Thanks!
@zerotousers We were also in similar situations -- Ruby on Rails, Apps to run and before the Heroku bill get skyrocketed we had to do something.
So we started building our own Heroku Alternative. You can read more about it here.
you can always keep a VPC and use Dokku to host multiple applications, easily. Dokku is essentially self-hosted Heroku
Thanks for the thorough review, Darko! As an indie founder myself, I've been searching for a Heroku alternative ever since they discontinued their free tier. Your detailed comparison of Fly. io against Render really helps shed light on the options out there. It's great to see Fly. io addressing the needs of budget-conscious indie developers, especially with their more transparent pricing and support for SQLite. The anecdotal evidence and your personal experience certainly make a compelling case for Fly. io. The downtime concern is noted, but given their funding and investor backing, it's likely they'll work on improving that aspect. Your insights are immensely helpful for those of us navigating the indie hacker landscape. Looking forward to trying out Fly. io and seeing how it works for my own projects. Thanks again for sharing your experience and thoughts!
Another option that I am currently working on and using to deploy my laravel and php applications is https://serverfluent.com - I have been building this as my hobby project for a while now. It provisions your servers and works as a small CD for automatic deployments. You can do all this stuff by yourself with something like ansible and some automatic deploy scripts, but not having to do this manually can be a huge time saver!
We're still using Heroku for our apps (been with them for over 10 years now).
We pay quite a bit, and I wish it was cheaper, or at least gave more for the price, but I don't feel alternatives are quite there yet.
I have gotten good help from their support on specific and important issues, despite not having subscribed to their $1000+ support option.
Heroku is my go to platform for hosting apps. I don't have an another platform to switch atm. Aws and google cloud is what i use for big scale apps so i can manage everything on a micro level
Due to the retirement of Heroku's free tier, developers have faced a lot of challengers in terms of node.js hosting services. When Heroku stopped the free plan, I searched many hosting services such as Glitch, Render, Railway, etc. There was another node.js hosting service named Fly. It turns out to be the most closest thing to Heroku.
I have tried using it, it mostly specializes in deploying docker containers and also the overall process of deployment is a lot easier
Thanks!
Will need to check it out for my next project!
What do you think of Vercel + Supabase ?
I'm using fly io for my app. It's a Phoenix framework (elixir) app and so far it's been pretty good. I've only just started recently so there isn't much traffic.
I wonder if reliability can be improved through having multiple machines in different regions?
This is really interesting, what's the feeling of something like Railway App (https://railway.app/)? I've been using them for a bit and their onboarding was lightning fast compared to something like Heroku though I've not heard what they're like in a production environment (if anyone has experience with them)
Even cheaper than this is running multiple apps on a $5/m VPS with the Piku open source PaaS. Installation is as simple as running a single script on a fresh VPS.
And if you need a hand installing Piku hit me up at pikuvps.com!
Fly is excellent and so simple to use.
I deployed some dockerized Rust apps without any problems!
I love fly.io, but their major drawback for me is how they handle logging. They store only a tiny amount of logs, and you have to connect to a third party to save all logs.
Is this a paid advertisement?
How do you compare with https://railway.app? It seems fly.io and railway.app offering is very similar?
Wow, thanks @zerotousers and @csallen for your insights! I was looking for a platform like this. Will give them a shot!
I have upvpn hosted on Fly for few months now.
Usage gets me billed only for production environment,
and staging environment runs under their free tier!
I usually add more regions when I'm launching upvpn on big platforms and expecting traffic, and then remove those extra regions when not needed.
They do a good job of documenting that its their Postgres is not managed database service . I rooting for them to bring managed postgres soon.
I have had no issues so far (touch wood, and keeping fingers crossed)
Thanks for the insight.
thanks for the research and for sharing your experience! Looks
This comment was deleted 10 months ago.