Hi, I am curious what do you think about Proxmox VE. Can plain Debian or other linux distro with installed LXD compete with Proxmox if we take as the main criterium functionality and security?
Proxmox and LXD both have features that the other does not. It’s worth comparing if you do not use certain features of one or the other.
For example proxmox can set up a ceph cluster for you and you’ll not have to think about anything. Or live-migration for VMs. Also proxmox is more plug & play. After installing you can browse to a web interface for management and deploy a VM right away. LXD requires some manual initial configuration for your network and storage.
LXD on the other hand is incredibly light-weight, can be installed on a distro/hardware of your choosing and is in my opinion way more mature. The docs are really complete. It is way more friendly on your resources, can easily run on a raspberry pi.
If you happen to find a bug and report it, it is very likely to be fixed the same day. If you request a feature it is likely to be put on a relatively short wishlist and added quickly. Proxmox on the other hand has thousands of open bug reports, some even 5 years old, likely never to see a fix.
If you prefer web management over cli you can use https://www.nuber.io/. Manually configure LXD to your liking on the cli and then use this sleek looking web interface to manage your local cloud.
Personally I have used both. I currently have several instances/clusters of both running at the moment. The only reason I have not replaced my proxmox clusters with LXD yet is because of the live migration feature for VMs. As soon as LXD gets that feature it’s bye bye proxmox and never to be seen again.
In my opinion LXD wins hands down
I’m currently working on a built-in DNS server for LXD but live migration of VM is next on my list after that. Though I have to warn you that it will initially be somewhat limited.
We’ll initially retain our host passthrough config, so if you’re migrating between CPU generations, you’re likely to hit some problems. We may also need further changes to our incremental migration protocol to make things efficient, so I suspect the first pass will be reasonably fast on shared storage (Ceph) but quite a bit slower on non-shared storage as we’ll effectively be doing something like a btrfs or zfs send/receive of the whole thing as part of it and may not be initially able to do multi-pass (so will need longer downtime than would be required with 2-3 passes).
But yeah, it’s something we’re quite keen on seeing work and something I expect we’ll be improving as we head towards LXD 5.0 LTS next year.