Interesting take. the PVE devs have made their position known on the subject, which you decided they "have neither the resources or the knowledge." Lets suppose thats true; why are you speaking as if that is some personal insult to you?
Because there are at least half a dozen tries behind me to help this to be at least moved forward, and almost every case ended up in silence, or with an example of why was it impossible, instead of something tangible like actually talking with criu people and actually trying to examine the problems, and maybe, just maybe letting the people doing criu development responding the problem instead of answering those instead of them from the sideline. And, as I mentioned, it's been going on since forever, and in the beginning it was "we're working on it, soon, next version maybe" and then silence for years and when I asked about it it was mainly… silence, or why-nots. I do not remember any case when they would show me any real discussion about the specific problems they mentioned. I got tired and annoyed, and nothing happened since to make me feel better. And I don't bother them about it since otherwise they're nice people, doing useful stuff, that's one reason they get a helluva lot of money from us. [Minor for them, maybe. Who am I to judge?]
Also, I have seen kernel development, and I have seen proxmox development, like, they can't even fix a broken scrollbar for 5 years now, because they used a broken external lib 5 years ago and they can't move over to a nonbroken one. I do not believe that if this is an unsolvable problem for them then a full featured kernel based checkpointing (which has been in development for 10+ years by criu) would be in their league. But I'd love anyone to prove me wrong.
Funny indeed. Why are you not using openVZ, but insisting PVE be turned into it?
Its always been POSSIBLE to migrate containers in a manner similar to virtual machines, but no way to plug the massive security holes involved.
Not in proxmox, no. Well, not since moving over lxc (which wasn't a choice, since OpenVZ was pretty much discontinued.)
For anyone serious, Kubernetes represents the modern way to do container clustering- plus many more features not even possible using lxc.
I understand your
opinion. K8s is good for some tasks and horrible for others. (Also it's a kludgy-buggy hell, but maybe I only say that because I looked at what's inside.) People choosing containers (lxc) do it for a reason and it's not the right approach to tell them to use something else you like instead. Same as with VMs, they're good for some tasks and very not good for some others, and they are definitely not lighweight.
I seriously dislike your categorisation, but I can live with that, and the world will survive (that at least) as well.
