exec should wait for shim

[dlespiau]

I don't really want to have to queue stdout/err data in the proxy.

To solve that, let's introduce a new constraint. The data path:

  shim <-> proxy <-> hyperstart

has to be fully setup before we allow newcontainer/execmd to execute a
new process. While we advise to start the shim as early as possible and
issue the newcontainer/execcmd command after that, we need to
synchronize somewhere. The easier way of doing that is to stall the
newcontainer/execcmd commands until we see the corresponding shim
registering itself.

Fixes: https://github.com/clearcontainers/proxy/issues/21
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

[coveralls]

Coverage increased (+0.7%) to 70.567% when pulling ab74862 on dlespiau:20170410-exec-should-wait-for-shim into 653860d on clearcontainers:master.

[jodh-intel]

This seems like a reasonable timeout, but since it is a relatively long time, I think it could be useful to display a message every second or so that shows the status. Something like:

Waiting for shim to register itself with token %s (timeout in %d seconds)

That way, it'll be clear from the log what the proxy is doing and that it hasn't just "hung".

[jodh-intel]

Second occurence of 20 * time.Millisecond - might be worth a variable for this?

[dlespiau]

PR updated:

• Added log message indicating we're waiting for the shim (didn't implement the "every 1s" enough, that's probably over doing it)
• Added a const for the 20ms. Actually not too convinced about that one, it's really two different constants that happen to have the same value, but ¯_(ツ)_/¯ (sorry I really wanted to use that emote!).

[coveralls]

Coverage increased (+0.4%) to 70.224% when pulling 20b364a on dlespiau:20170410-exec-should-wait-for-shim into d3321c9 on clearcontainers:master.

[coveralls]

Coverage increased (+0.4%) to 70.224% when pulling 20b364a on dlespiau:20170410-exec-should-wait-for-shim into d3321c9 on clearcontainers:master.

[jodh-intel]

lgtm