This fails for me on xenial after upgrading from Docker v1.13.1 to 17.03.2-ce from xenial-proposed. Here's what I did:
1. Deploy a Canonical Distribution of Kubernetes cluster (which deploys worker instances with Xenial and gets Docker v1.13.1 via `apt install docker.io`)
2. On all worker units: add xenial-proposed line to sources.list.d, then `apt update && apt install docker.io`
3. Run `docker --version` on the worker units to confirm 17.03.2-ce was installed.
At this point, I observed that Kubernetes could no longer create containers, with errors coming from Docker along these lines:
Create container failed with error: invalid header field value \"oci runtime error: container_linux.go:247: starting container process caused \\\"process_linux.go:334: running prestart hook 0 caused \\\\\\\"fork/exec /usr/bin/dockerd (deleted): no such file or directory\\\\\\\"\\\"\\n\"
After observing this failure repeating for several minutes, I did the following:
4. On all worker units: `service docker restart`
This fixed the problem: the fork/exec errors from Docker stopped, and Kubernetes was able to create containers again.
This fails for me on xenial after upgrading from Docker v1.13.1 to 17.03.2-ce from xenial-proposed. Here's what I did:
1. Deploy a Canonical Distribution of Kubernetes cluster (which deploys worker instances with Xenial and gets Docker v1.13.1 via `apt install docker.io`)
2. On all worker units: add xenial-proposed line to sources.list.d, then `apt update && apt install docker.io`
3. Run `docker --version` on the worker units to confirm 17.03.2-ce was installed.
At this point, I observed that Kubernetes could no longer create containers, with errors coming from Docker along these lines:
Create container failed with error: invalid header field value \"oci runtime error: container_ linux.go: 247: starting container process caused \\\"process_ linux.go: 334: running prestart hook 0 caused \\\\\\\"fork/exec /usr/bin/dockerd (deleted): no such file or directory\ \\\\\\" \\\"\\n\ "
After observing this failure repeating for several minutes, I did the following:
4. On all worker units: `service docker restart`
This fixed the problem: the fork/exec errors from Docker stopped, and Kubernetes was able to create containers again.