Checkpoint and restart recovery

[containerd]

Checkpoint and restart recovery

[Random-Liu]

@yanxuean Are you interested in this? We could start from a simple PR, just put container/sandbox metadata into container label when create container.

[yanxuean]

@Random-Liu Ok.

[Random-Liu]

Just FYI. We could leverage containerd/containerd#1378 to store our metadata.

[abhi]

@yanxuean thanks for offering help with this. Let us know if you got a chance to take a look at it and if you need any help from us.

[yanxuean]

OK. I'm studying the RunPodSandbox function now. If there is a need I will.
@Random-Liu Do you means that we should put sandboxstore.Sandbox into containers.Container.extensions by json string?

[Random-Liu]

@yanxuean sandboxstore.Sandbox.Metadata and sandbox.Container.Metadata.

And if we use the new extensions, we don't need to convert it to json. Just put the type into Any direclty.

[Random-Liu]

@yanxuean How's going? The first step is relatively straight forward, but this feature itself is important to this release. :) If you haven't got a chance to look at this, we have to start working on this now. :)

[yanxuean]

@Random-Liu I modified it, and I'm testing it, but it always is fail.
because containerd/containerd#1378 has not commited, I merge it on local(api/next.pb.txt, containers.proto, container.go), and add withextension function.
I'm analyzing the reason.

[Random-Liu]

@yanxuean Could you add them to container labels for the first version? So that we could start consuming that information.

After containerd/containerd#1378 is merged, we could leverage it, but the restart recovery logic won't be affected. :)

[yanxuean]

@Random-Liu OK, I modify it today

[Random-Liu]

@yanxuean Thanks!

[Random-Liu]

Reopen since this is not finished yet.