About workaround on bug of runc

11/21/2016

Related to the following URL: Kubernetes: runContainer: API error (500): Cannot start container (docker failed to umount)

We are using the kubernetes cluster on GKE. The version of kubenetes is master version 1.4.5, and minion node version 1.2.0.

The same problem as the above URL occurred on our cluster. We understand that upgrading node version is one solution, But currently it is difficult to immediately upgrade the minion node with node-pool due to system restrictions

So I have a question. In answer to the URL above, CJ Cullen replied as follows

You may be able to work around the issue by adding extra characters to your container's command line.

In kubernetes running on GKE, what should be changed to apply the above workaround? Should I change the manifest file or Dockerfile?

-- umiyosh
docker
google-kubernetes-engine
kubernetes

1 Answer

11/21/2016

Adding extra characters can be done by adding an environment variable with just spaces as its value (or with no value in most cases). Please note that this won't be a full fix -- because the bug occurs when your container config is a very specific size (a power of 2) and you can't be sure that adding extra characters will not cause your container config to hit the power-of-2 value.

-- cyphar
Source: StackOverflow