pause: Remove all things related to pause container

[sboeuf]

The pause container was introduced long time ago because our agent was not able to maintain a pod running when the last container had exited. But now that our agent has this capability, we don't need to maintain this workaround. Moreover, this will simplify our code base.

[coveralls]

Coverage decreased (-0.1%) to 65.889% when pulling c6513bdc8a5e08727dcd933af9fe4571e0f37d9b on sboeuf/remove_pause_container into e3d41a0925a90e5704ad9fe46618da2fbba2726f on master.

[jodh-intel]

\o/

lgtm

[jodh-intel]

@sboeuf - could you handle the removal in the runtime too when you revendor?

[jodh-intel]

... aaaand I've just come across https://github.com/clearcontainers/runtime/pull/756 :smile:

[devimc]

magically cc-agent now is able to maintain a pod running? how? commit message is not very clear, would you mind explain a little more how was that possible?

[sboeuf]

@devimc well basically our agent was always able to run without this, that's a legacy thing we had introduced when our former hyperstart (the one from Hyper) was not able to maintain a running pod with no containers running.

[devimc]

@sboeuf make sense, thanks

[sboeuf]

@jodh-intel @sameo do you mind merging this PR even without the CI passing here. I mean this is not worst than before, when we're going to vendor this from runtime, we will see if something got wrong, and we won't break the runtime. But of course, I have submitted https://github.com/clearcontainers/runtime/pull/756 and the CI is passing, so I think this is covered and prove nothing is gonna be broken. I am asking to avoid blocking our PRs while we find a proper solution for the case where virtcontainers API or exported structures change.

[sameo]

LGTM