> This would be a lot more compelling if it included testimonials from actual projects bitten by any of the arguments against docker.

Disclaimer: I am not affiliated with the author. I just experience endless issues with docker in production. Enough to write a book. (just testing future audience for the book)

That, I can provide in a separate blog post! :D

What selection of bugs would you be interested in?

- Critical linux kernel bugs killing the host when running docker

- Containers crashing and corrupting all data on the mounted directories [from the host drive]

- Worldwide outage of docker.com killing all CI pipeline in the world for 7 hours straight

- More critical linux kernel bugs destroying the docker host

- Major breaking changes in docker and docker API (usually between all minor versions), breaking all existing stuff

- Randomly dropping critical features, distribution and filesystem support (to the point where there is actually NO filesystem supported by Docker on some distribution. No kidding xD)

- Various DNS fuckups (special issues on Ubuntu)

- No build-in commands to free disk space of old containers/images (this one is major hassle)

Awesome, you _should_ write a blog post if you have time :)

I'm particularly interested in what you did to get around all the bullet points. Are you still using Docker, with some bandaids, or did you move to something different?

Hope you're still here. Comments and feedback welcome.

https://thehftguy.wordpress.com/2016/11/01/docker-in-product...

Thanks :)

Isn't Nix(OS) ready for production? As far as I know it does the same as Guix, but without LISP.

Sounds like it, I don't know enough about Nix to know if it addresses the same scope of need that docker does, hopefully someone who is savvy can chime in :)

It's "production ready" if you're willing to do a lot of work yourself. It has a lot of solid theoretical and practical advantages, if you're willing to do so -- as well as some downsides. Advantages include things like real reproducibility, a "unified language" that can describe everything from your packages to server networks, atomic upgrades and rollbacks. More minor advantages are things like powerful development environments, automatic support for distributed/remote builds for your software/deps, easily supports customized software (it's very easy to patch your kernel and have all kernel dependencies rebuilt automatically, stuff like that etc), and other stuff. It's probably one of the most forkable/customizable distros, just purely due to the way Nix works, and fact everything is in one git repo, making 'wide changes' as easy as a branch. Downsides include the fact that the security story is still ongoing on several fronts (userspace/kernel hardening, quickly shipping userspace vulnerability fixes) and that compiling everything -- when it happens -- can take some time. (I only list these two, but obviously these aren't small problems and IMO are two of the biggest ones).

NixOS certainly isn't ready for "ordinary consumers" or anything, even among people who might use Linux regularly. It also just factually isn't going to be to everyone's exact taste, in terms of some design decisions (hence Guix exists, and Guix is great!) But for advanced users and operators, it can certainly be well worth it, and confer some unique and powerful advantages.

Note that I'm specifically skirting the question of "does it address all the same things as Docker" (in fact, they aren't even fully competing projects, in many ways) -- because that's a fairly complex question given all the ways people use both projects, and it could fill a blog post (or several) in its own right.

Awesome, thanks, this is really helpful.