Hibernating a Pod????

[stellarpower]

Hi,

This is a bit out-of-the-box, but what would people think about the concept of stopping a pod and saving its in-memory state to disc, similar to hibernating a machine?

A usecase for servers might take more thinking, but I am coming more from the angle of devcontainers and the like in my motivation. If one is working on a particular feature, and then something comes up, it would be a nice way to be able to put that to one side, shut it down in a way that it can be restarted again, and reclaim any system resources. Every time I have enabled a swap it ends up just locking up the kernel quicker as it desperately tries to flush things to disc only once it's too late. As a devleoper who has to change between unrelated projects on a regular basis, and ho normally segregates projects into containers anyway to manage dependencies, I am thinking with this option, podman might be a good way to switch tasks without having to keep everything open all the time.

How technically plausible or inplausible would such a thing be, and can anyone else see any very different usecases for something like this?

Cheers

[Luap99]

well podman pause is a thing a to freeze the processes so this will save cpu but not memory AFAIK.

There is also podman container checkpoint which uses criu to actually save and serialize the process state. The problem there is that is a privileged operation that only works as root. So you cannot checkpoint rootless containers with that and it also does not work for all processes/containers.
Generally I would recommend against it for such a use case.

The simple answer should be just stop and start the containers again when you need them?

[baude]

the lines between containers and vms has certainly become blurred. to me, in memory capture of some process is more of a vm thing and kind of bucks the original intention of containers being ephemeral. that's just one old timer's view of the container world.