Deployments after the retirement of Heroku free dynos

Hello,

Ever since picking up django in 2020 I have been deploying projects with Heroku. With the announcement last week of the retirement of all free dynos I was wondering if people could share their deployment providers, pricing and ease of use.

I have 5 - 10 small projects that do not justify the pricing for the dynos for each project on heroku. It would be a shame if I have to retire these projects.

Any feedback would get greatly appreciated.

Leon

I’m late to the party on this post; I’m curious what you ended up doing and how that went for you, if you have a moment to share.

Are you still on Heroku? I have some recs on PaaS if you are looking.

No, I’m not on Heroku. Ultimately I’m leaning toward a plan to try out Hetzner by itself and then potentially try out using Coolify or Appliku with it, but I welcome the recs / thoughts

I was researching PaaS alternatives beforehand, and what I found was that the options I saw are really fixated on AI right now. I have ethical and economic concerns around AI, so right now I’m pretty resistant to choosing a platform that has it incorporated in a meaningful way. I did try Upsun a few months ago despite these reservations because of the lack of PaaS options that aren’t betting heavily on AI plus it’s a certified b-corp plus I thought I’d be able to use django-simple-deploy for it (but I found out the hard way that Upsun had changed making its plugin for django-simple-deploy out of date). I still tried deploying with Upsun with some degree of success, but for [reasons] I stopped using it; it’s possible I’ll circle back to it if I can’t figure out the Hetzner option.

Sidenote: I watched both versions of your deployment talks this year, and I think your description of why this landscape is confusing/intimidating was spot on. I know the purpose of your talk was to discuss deployment at a level above the vendor specifics (which was great, and I keep pulling up that summary slide at the end to situate myself), but I also have to complement that slide you had in the EuroPython version with the logos categorized into types of vendors (e.g. VPS, PaaS, etc.) – organizing it like that was helpful for my understanding plus I discovered a couple of options I hadn’t come across before.

Well thank you. On AI… I’m in the middle of it with my day job at JetBrains and my own work. I applaud and agree with much of your anti-AI take, but I think choosing a hosting provider based on that is going to lead to short-term frustration. They are ALL going to try to adapt to an agent-first world whether that becomes the default or not.

I’ve heard good things on Hertzner, especially with Coolify or Appliku. I would definitely go for a managed database. Personally I’ve been trying out Laravel’s Cloud new Django/Python offering and it’s quite good, but Vercel also has new Python stuff.

Honestly, it’s hard to choose. I would pick just something and know that you can pretty easily switch unless your website is super-high traffic in the future as needed.

Hello, what a pleasure to have you both reply to this topic. I’ve been part of the Django ecosystem since 2020, so I’ve seen your names around here and there.

It’s been a bit of a journey over the past couple of years. In the end, I did migrate some of my personal projects away from Heroku, while others had been dormant for a while as I evaluated the next steps.

Personal projects

I managed to get Django running on Railway.app, and their Discord community was very helpful. After a while, however, they also retired their free tier. To my knowledge, this has since been brought back, although the last I heard about it was around three or four months ago.

Then there was Fly.io. I had things running there, but while hosting my database on the platform, Fly.io experienced a hardware failure. I didn’t have a replica, unfortunately, which resulted in data loss. While this was partly my own fault, and I acknowledge that, I was fortunate that it was a personal project rather than something with users depending on it.

Some of my projects could also be moved away from Django altogether. These were typically single-page sites where I could get away without a database, for example. I moved those to Cloudflare Pages, usually using Nuxt or Hugo, depending on the requirements.

Work-wise

I work with my brother-in-law, who at the time was starting to get more involved in DevOps. We tended to stick with Heroku for work in the meantime, largely because we were familiar with it. However, we also looked at AWS, particularly CloudFormation, as we liked the idea of infrastructure as code.

With a bit of Elastic Beanstalk and RDS, we could get our Django projects running quite nicely.

With further research, we came across Terraform around three to six months ago. This was great, essentially CloudFormation on steroids. Being able to integrate with external providers such as Cloudflare was particularly useful, as it meant we could describe an entire project, from DNS through to the database, in code.

During this time, we’ve been (and are still) migrating our Django projects from Elastic Beanstalk to Fargate. This allows us to put environments to sleep when they’re not in use.

We still have a few things left running on Heroku, but certainly fewer than before as we work through the list and migrate them over to AWS.

So, to answer your question: Fargate, RDS and the wider AWS ecosystem, using Terraform.

This is how my brother-in-law and I are currently deploying Django.

We’ve also had a look at Cloudflare’s recently announced approach to deploying Django, but it’s not quite ready for our production environments just yet. I look forward to seeing how it develops.