Thoughts on self-hosting & PaaS (May 2026)
Right now I’m running Practical Computer’s apps on Heroku, based on the following criteria:
- Crunchy Bridge is, IMO, the best Managed Postgres option out there. Their platform offerings mean I’m locked into being in AWS’s network to avoid latency hits. I learned that one the hard way with the performance hit of crossing datacenters between Hetzner + AWS when self-hosting via Dokku ( microblog.thomascannon.me/2024/05/1… )
- As other folks have said, my time is valuable, so I want minimal babysitting of servers
- Heroku’s developer experience is still really solid given those constraints
- Render has real “Ambulance Chaser” vibes that put me off
However, given those constraints, I’ve still done a bunch of research into what self-hosting would look like, especially for the extremely tiny apps that aren’t self-sustaining. The criteria I have for self-hosting are:
- Container-first (which disqualifies Hatchbox). You can’t beat the portability & isolation of container images
- Sustainable & wide community support outside of the Ruby/Rails ecosystem. This disqualifies ONCE/Kamal, which are “Not Invented Here” projects
- Have strong Infrastructure As Code foundations, notably Terraform/OpenTofu support. Terraform/OpenTofu support is critical in my eyes because if I’m giving up the developer experience of a platform like Heroku, I want to make up those time savings elsewhere in the administrivia of provisioning and maintenance
Right now the current candidate is Portainer (www.portainer.io, registry.terraform.io/providers… ). And even then, it’s an annoying lift that I’m not making anytime soon.
For a more managed alternative to Portainer, I’ve also been looking at Northflank, but am not biting the bullet there yet: northflank.com
