Hi, all, I am conducting my master’s research in current methods to perform live migration on containers. I have found that all migration techniques are heavily dependent on criu. As it turns out using Ubuntu 18.04 as host os, I have been unable to live migrate anything using lxc move, except alpine containers.
Keep getting several errors associated to criu in all cases.
Even with alpine the moment I install an SSH Daemon it fails. Can you suggest any scenario when live migration actually works? At the time of writing I extremely doubt that criu is operational at all. If can’t migrate a plain Ubuntu container, what is the point of lxc move at all?
CRIU indeed does need a lot of work in order to be functional with common container setups.
The majority of our active live migration users are working with workloads that are particularly designed to work with live migration. This typically involve unloading/unconfiguring anything that CRIU cannot handle (be that apparmor profiles, some network settings, …). Then ensuring that you don’t have any lxc exec sessions into the container as those also can’t be dumped and lastly reduce your process list to the bare minimal you need (likely your own software) and similarly, make sure that this code doesn’t rely on anything CRIU can’t dump/restore.
This does make this unsuitable for most users and is why CRIU is disabled by default in the LXD snap, but for those few who particularly tailor their environments to live migration, it can be a useful tool.