Remind HN: Heroku will delete all free dbs and shut down all free dynos Monday

Now that Heroku free tier is not available, here are a few alternatives

Serverless and edge functions: Deno, Vercel, Netlify, Cloudflare.

Databases: Planetscale, Supabase, Cockroach, Fauna, Upstash.

Services: Fly, Railway, Render.

https://mobile.twitter.com/Ak_Mittal/status/1567175784229650...

To be fair, I have had quite a few emails form Heroku telling me to move my stuff. Even after I thought it was already shut down, I keep getting emails. Everything I have on there is 5+ year old hobby things that I won't bother hosting elsewhere, it's nice to know they are paying more attention than I am to my projects' fates

I have had quite a few emails form Heroku telling me to move my stuff.

Heroku's warning emails are how I found out two things:

- That I'm still the administrative contact for a former employer's production environment.

- That my former employer is using a free tier for its public-facing web site.

Ethically, I should let the company know is web site is about to go offline. But it's a start-up run like a fiefdom and the founder tells all departing employees that if they contact anyone at the company, he'll sic the (imaginary) lawyers on them.

A fun fact (and I'm not violating any NDA here) my client https://paireyewear.com is growing very fast and last year they raised $73 million from VC investors. They have their main store on Shopify, so they didn't need too much on the backend: only their CRM for keeping customers happy and handling returns and damaged goods and refunds and special sales. Also, on the backend, they had all the integrations with the 3PL, that is, 3rd Party Logistics, which is to say, they worked with external warehouses, because they weren't ready to have their own warehouses. They did have integrations with the warehouses so they could track inventory levels.

So the fun fact is this: they were running entirely on free Heroku dynos. This year they successfully transferred to AWS, but they got to a fairly big scale while running entirely on free Heroku dynos. I'm still kind of amazed by that.

> they were running entirely on free Heroku dynos. This year they successfully transferred to AWS, but they got to a fairly big scale while running entirely on free Heroku dynos.

I’ve read so many of these stories now that I understand why Heroku is phasing out the free tier.

It’s kind of fascinating to see how some engineers see free tier limits as an optimization target. I wonder how many engineering hours across the industry went into arbitrarily keeping services small enough to fit in the free tier.

> I wonder how many engineering hours across the industry went into arbitrarily keeping services small enough to fit in the free tier.

Why is that wrong? If everyone optimized their services and applications like it used to be in the past, will substantially bring down compute and memory requirements and reduce overall bloat.

It’s not, strictly? It’s just more of a tragedy of the commons.

Heroku is not "the commons", they're a company and people are playing within the rules that they set out.

This has to be the analog of "the first amendment only protects you from government restrictions on speech" in this conversation.

Heroku can protect itself from abuse of its free tier by changing the rules - which they eventually did.

I'm sure that the only reason why they kept the free tier as long as they did was because it brought in enough users (who were eventually converted into paying customers) to be worth the cost.

> I’ve read so many of these stories now that I understand why Heroku is phasing out the free tier.

To me, those stories seem like good arguments for reducing the usage limits of the free tier, but not getting rid of it entirely.

You don't need to profit off every single customer. Many business models make 90%+ of their revenue from 'whales' and crumbs from the rest. Optimising for consistent margin from every single customer could be a waste of time; a distraction, and could even hurt a business.

How could they stay on free dynos? You can't be running on a free dyno 24/7. It's like 16/7 or something.

That was at one point true, but hasn't been true for at least a few years.

Heroku could have gone back to that i suppose, instead of getting rid of free dynos entirely, but they didn't.

Still... I think you could basically only run one free dyno per account 24/7. I am curious if they somehow fit their entire production in one free dyno, or were somehow splitting things between multiple accounts, or what.

I have several legacy free tier dynos that don't have the restriction. They run 24/7, at least for another day or so.

If I'm not mistaken, if you gave Heroku a valid credit card, it would add a few hundred extra hours for your dynos.

that's correct, you could be have it on 247 with trick!

Shopify does logistics now too, you could use that and get rid of the whole backend altogether. Disclosure, I work there.

They’re part of the reason it’s not free any more.

Well, Salesforce could have made the free dynos non-commercial only

That's even worse, so when your hobbyist project accidentally makes 40$ Salesforce can send it's legal team after you?

I would prevent any hobbyist project of mine from making $40 for tax reasons alone, where the difference the to tax authorities between $0 and $40 is basically the same as the difference to a Saleforce non-commercial clause. At $0 I don’t have to do anything about taxes. At $40 it’s immediately in the red due to whatever time I have to spend sorting out taxes for $40 in revenue.

Cool story (really), but what were they hosting on Heroku then?

Thanks for that reminder. I had to delete an app I had running since 2013. Up until the moment before I committed to deleting it, I felt no emotion at all. But right after I confirmed deletion, I felt like an animal I had on the farm died. I was sad. I had that thing chugging along with no real issues for almost 9 years. I used that platform to learn and try new things and then built an app to track my workout progress. It was a great way for me to learn Django without losing money and build some useful tools.

Have they committed to anything different than what Heroku had committed to?

We're Charlie Brown running towards the football repeatedly. We need to learn a lesson here. "We're the good platform devs" branding doesn't mean a company will do the kind thing by their true believer users (post-acquisition, post-cash-flow-belt-tightening, post-IPO, etc)

There ought to be a platform on which new coders can build a free simple web app in a playground, and know it will be accessible in 20 years. If a company wants to use "free easy backends" as bait for capturing growing companies' future costs, devs should hold them to long-term persistence and at least an off-platform future migration path. See also Parse, Geocities, etc

Use some % of the money and put it in to an OSS migration path or a fund to pay their own future AWS bills, c'mon.

Why are you expecting 20 years of free service?

Even if it's a penny a day, that's 60$ over 20 years supporting a user with no intention of spending any money.

If you still have the code, you can always deploy it again.

People need to realize that free plans are a loss leader/sales feature, not a guarantee. The longer you're able to use their free plan, the more obvious it is that you aren't going to convert to a paid user. The better the free plan, the less likely it is to make sense to run long term.

The only free tier product I use is Cloudflare's, and if it goes away someday, I'll pay for it.

Many are new students who don't think about the business landscape of the service they choose to build something on.

Students who are learning to build things on these platforms shouldn't lose their early work.

Sure, but move it elsewhere later. And who doesn't keep a copy of the code they deploy?

> Students who are learning to build things on these platforms

... and have not yet learned about version control or backups the hard way.

It’s a good thing those students have access to thousands of dollars of free credit for AWS/Digital Ocean/Azure, etc.

Not to mention if whatever they are working does take off, those companies would throw extra credits to these students.

I think they’ll be fine.

Well, then it's excellent they are learning this now and not later. The best time to learn about making backups is when you start programming, the second best time is today.

Glitch's free plan is pretty much useless for everyone except the intended audience: new programmers looking to get help from the community (there's a button) and experienced programmers looking to provide that help. It has tiny resource limits. $8/month lifts some of the limits for all your projects and gives you five "boosted" projects with more memory and storage.

People who've been around a bit will recognize its previous name: Fog Creek Software.

> We need to learn a lesson here

Something comes to mind about a free lunch?

The idea that there ”ought” to be a free service that guarantees you 20 years of service is laughable.

If you’re paying nothing, guess what you can demand?

There's plenty of platforms where devs can do that. The idea that there should be One Monolithic Platform that lasts for 20 years is silly. Why only one? We have tons of options from modern hosts who have some kind of free plan available.

There's very little incentive for hosting companies to offer grandiose 20 year perfect free backend plans because hosting has been extremely commoditized and anyone hosting on top of AWS is just being squeezed aggressively. While you waste 25% of your resources fighting bad actors, spammers, CSAM, and everything else that comes with being The Place for Free Backends, your competitors are staying lighting focused on the features that your paying customers are leaving you for, spending only the minimum necessary on free options.

Frankly, if a dev can't find one of the ~dozen free top tier amazing backends available to newbie devs (many of which available through entirely web based IDEs that can launch any framework you want in moments), maybe they're not ready for this field, because a lot of it is in fact their ability to search and learn.

do they only offer JavaScript backend? I am looking for other language support

Just moved off my 8 years old project google-webfonts-helper https://github.com/majodev/google-webfonts-helper from their free tier to my own private infra and replaced the current dyno with a 301 handler: https://github.com/kenmickles/heroku-redirect

AFAIK sadly Heroku does not provide some other _free_ permanent redirect option for their *.herokuapp.com sub-domains without actually running a dyno there.

For alternatives we can recommend Hetzner US Cloud [1] which was the best US Cloud provider we've found that works out to be an order of magnitude less expensive than equivalent specs on Azure/AWS and also includes 20TB free bandwidth that would cost a fortune in AWS/Azure's artificially inflated egress costs [2].

The UX behind managing instances is delightfully pleasant where new instances are available faster than any other cloud provider we've used, within seconds of creating an instance you can immediately login with your configured SSH keys. Another nice feature is being able to "rescale" your instance to higher specs after a restart [3], so you can confidentially start with a small instance that just meets your current workload knowing that you can easily scale up your instances as your workload increases.

AWS RDS was the only critical service keeping us on AWS, a service we no longer need in our new Apps which we're building with SQLite thanks to the effortless replication in Litestream [4] that we're using to replicate to Cloudflare R2 - another great value alternative S3 alternative with $0 egress fees [5] where you can get even greater value & performance when hosting behind their free CDN.

[1] https://www.hetzner.com/cloud

[2] https://servicestack.net/blog/finding-best-us-value-cloud-pr...

[3] https://bizanosa.com/how-to-upgrade-resize-hetzner-cloud-ser...

[4] https://docs.servicestack.net/ormlite/litestream

[5] https://www.cloudflare.com/products/r2/

Thanks for the reminder! Could've lost a bunch of photos :) Also, it's interesting how little effort Heroku put into helping you upgrade. There's no one-click options, and instead it's a forest of choices and menus to jump through.

By the way, one hack to not upgrading your app? Move your DBs to a single shared instance on Amazon, and then point Heroku apps to that one instance. Meanwhile, in Heroku just remove the Postgres addon. As long as it doesn't see a free instance, it doesn't complain.

Looked into it a few seconds more and realized they've got a new "mini" plan, which is basically a free Dev plan, but costing 1 cent a month. I suppose this is a deliberate attempt to kill all the non-maintained projects, rather than a way to make money.

Where are you reading 1 cent/month? The cheapest way I can see to run a small app+db project on Heroku now is to use an Eco dyno at $5/month with a Mini Postgres instance at $5/month, for a total of $10/month.

If you don't want your project to sleep, you probably want a Basic dyno ($7/month) instead of an Eco dyno.

Am I missing something?

Edit: I meant to add a link to the pricing page.

https://www.heroku.com/pricing

As of now, https://elements.heroku.com/addons/heroku-postgresql says $0.01/month and their CLI as of Wednesday also said that. Of course, this feels like a weird workaround until December 1 since they also say that's when they're going to saying charging for it.

Just found this too. The brand-new Eco plan costs $5/mo, shared over all projects up to 1000 dyno hours, and then it seems each database on the Micro plan is $0.01 per month. That comes out to $5.05 per month for me I believe for my 5 portfolio projects? Maybe there's hope for Heroku yet.

I can't find anything about the Micro plan. Is this documented anywhere?

Does anyone have a link to a good analysis about the "death" of Heroku? Say, a good Ben Thompson style piece explaining how it all was subsumed into Salesforce and failed as a standalone service.

I’ve migrated an old PHP7 app from Heroku to GCP the last couple of weeks. I rewrote the CRUD in Go. The rest of the components were an SQLite database and an old (2013) SPA that holds up surprisingly good even today.

Heroku account: deleted.

For small scale projects Heroku seemed thousands of times easier to work with than GCP. I sometimes get lost in all the GCP menus and settings. It’s a lot.

Last time a post about Heroku was posted, I talked about my great experience using Coolify:

I am using Coolify (https://coolify.io), an open source self-hosted PaaS which is a relatively newer kid on the block compared to Dokku and CapRover. I tried both of these and I just didn't like how they were, always had some problem or another.

In contrast, Coolify has a great GUI that abstracts away the most common things about PaaS hosting, like connecting to GitHub automatically for git push deploys, SSL certificates, reverse proxying and custom domain support, and best of all, having support for Heroku style buildpacks as well as Dockerfiles. I've been quite happy with it, the creator has a Discord and responds to issues very quickly.

With regards to non-self-hosted options, I did try out Render, Fly.io and Railway but I found that their free servers were too anemic. I was compiling a Rust backend and it simply could not compile on their free servers. On Hetzner, for 5 bucks I could get a 2 AMD vCPU and 2 GB RAM machine that was sufficient to compile my Rust apps in a way that the non-self-hosted ones were not. I have a JS frontend app that works fine though but I wanted to keep everything under the same VPS, plus I can run other types of self-hosted services on it too, like Plausible analytics and a Ghost blog. I'm not sure if those are allowed on non-self-hosted options.

All in all, it costs me 5 bucks a month, and I never have to worry about sudden upcharges for traffic à la AWS as in the very worst, my VPS goes down for a while. I'm now running about 20 different services on this 5 dollar box including databases and applications as well as other services, works just fine.

Hesitate to tell that Fly is the only good option since they won't hibernate your instances, and have a reasonable free db and bandwidth quota .. unlike Render (almost not count as free plan, more like some free trial)

I'm fearing folks could abuse Fly and they could start to let it go..

What did coolify have that caprover didn’t?

Better UI, more streamlined GitHub integration (I could not for the life of me get CapRover's integration to work), Docker compose support

So basically anyone with a bunch of free dynos (who doesn't use heroku DBs) can just switch to the $7/month eco subscription and keep their dynos running?

This thread was fine with me until I considered Vercel might do the same thing in five years with their free tier.

I understand this is news for a few months now but still, such a hostile way to just unplug projects and to an extent, the history of the web.

I doubt Salesforce is struggling with cash not to let these projects stay online on a “read only” mode.

What would you propose they do differently?

The single biggest thing would be to suspend the websites but NOT delete the underlying data - for at least a month, ideally six.

That way all of the people who missed the news about Heroku and first heard about it when sites they were using stopped working would have a fighting chance to recover their data before it vanished.

Even better: Salesforce could have actually invested resources in Heroku (instead of effectively feature-freezing it a few years ago) such that the thing where the free tier acts as a major funnel for attracting paying customers continued to work, as it had for most of the lifetime of the service.

Completely agree. Scale down to zero and turn off (access to/entirely) databases but do not purge any data for 6m or so, hopefully one last chance and enough time for someone to notice and go "oh no!" and pay up/export data.

Just shutting it down and deleting after notifying via email (did notices go to junk? etc?) is kinda shitty to do.

If you have a Heroku instance you definitely have a local copy of your code stashed somewhere.

some projects are +10 year old. People will go through multiple laptops in that time.

if your project used heroku directly without any github integration then they effectively were also your source control.

The PostgreSQL data is a much bigger problem than the code.

Does anyone have a quick and dirty script to download the Git repo and data for all your Heroku apps?

I’ve just been going through and cloning the repos via their command line tool:

    heroku git:clone --a <app name>

If you have really old apps (e.g. on the bamboo stack or older), there is a plug-in that allows you to download the slug (as Git doesn’t work with those old stacks).

I hope no one is rushing over the 4.5-day US holiday weekend to do a last-minute migration or backup.

To be fair, Heroku has been sending reminders for this every week for at least a month.

They can delete my dynos but they can't delete all the emails over months and months telling me they are going to delete my dynos. I'll never get that time back.

I kid.

Thanks for everything, Heroku.

Heroku will keep going alive and well. It will be integrated into many Salesforce services and the user base will come from those integrations.

It’s curios like this action push user to go way instead just paying 7$/month.

I think for many people there is a clear line between free and not free, no matter whether the service costs 2$, 7$, or 100$. It's a psychological thing. On top of this, it could be 7$ now, 14$ next year, and over time you get sucked into this money pit.

A good example is Adobe products. Pre-cloud I would freely use PS/Lightroom off a CD, then got into the monthly plan for students, which turned into a full price plan after I graduated, and which sucked me into a year long contract for a certain plan.

Now I did this all willingly of course, and after 5 years I looked back and thought 'where did all that money go?' But to some degree it undermines my trust in my ability to gauge the actual value of subscription based services.