- cross-posted to:
- linux@programming.dev
- technews@radiation.party
- cross-posted to:
- linux@programming.dev
- technews@radiation.party
Interesting move by Canonical. Wonder if this is related to the new GUI for LXD that Canonical released recently? Or maybe they want to bring more projects in-house after the RHEL shakeup?
I guess it depends in your use case. If you’re taking about “regular” applications LXD/LXC might not be your best fit. LXD/LXC seem to very good for the more low level infraestruture related solutions. In contrast, whatever is typically deployed with k8s that is mostly immutable very reproducible and kind of runs at a very high level.
LXD is more about what might power that “higher level” layer, more about mutable containers, virtual machines and very complex stacks that you can’t deploy with docker most of the time. As excepted people with those needs greatly leverage cloud-init and Ansible in order to get the reproducibility and the automated deployment capabilities that the Docker “crowd” usually likes.
Ah, ok, understood then, it didn’t fit my use-case or workflow, it works for others, my bad, appreciate the correction!
Not a correction, it has its uses :) I would never deploy a web app and its API, database etc. using LXD, makes no sense, k8s is way better for that.
I do that with lxd, but I have written ansible playbooks (almost like dockerfile? ) to automate the lxd containers. You could probably write some automation for scaling as well, but not something I’ve done, I have just opted for high availability with ceph & keepalived. Whatever works for your use case :) I do use some docker, but this is still nested inside lxd…
I also do playbooks to deploy stuff some stuff with LXD, but my end users only like Docker so, I kind of setup the infrastructure that allows them to deploy Docker on top of LXD containers that are deployed using Ansible.