This starts happening after #3195.

It seems that this completely broke the containerd Kubernetes e2e test:

• https://k8s-testgrid.appspot.com/sig-node-containerd#containerd-node-e2e
• https://k8s-testgrid.appspot.com/sig-node-containerd#containerd-node-e2e-features
• https://k8s-testgrid.appspot.com/sig-node-containerd#containerd-e2e-gci
• https://k8s-testgrid.appspot.com/sig-node-containerd#containerd-e2e-ubuntu

When testing locally, I get time="2019-04-10T10:17:03.792737608-07:00" level=error msg="post event" error="rpc error: code = FailedPrecondition desc = failed publishing event: namespace is required: failed precondition".

I'm not sure whether we actually support context metadata in ttrpc server.

• If yes, we must be missing applying namespace somewhere;
• If not, and if the ttrpc change is big, we may want to revert Use ttrpc in the daemon for event publishing from shims #3195 for now, add the support into ttrpc, and resubmit the change.