kubectl: exec errors with a segfault when the kubelet's runtime is rkt

[cgonyeo]

I spun up a libvirt coreos+k8s cluster that used rkt for its runtime, and kubectl exec would error out with a segfault every time it was used, regardless of if the requested binary existed or not.

derek@rokot ~/git/kubernetes> kubectl exec frontend-gbecy ls
error: error executing remote command: error executing command in container: signal: segmentation fault (core dumped)

The changes that I made to the cloud configs such that it would be a rktnetes cluster are here.

cc @yifan-gu

[euank]

I can't trivially reproduce this. Can you verify that you're using a kubectl binary that is compiled from the same commit? I'm using ./cluster/kubectl.sh -it exec my-single-container-pod echo hi and seeing a perfectly happy response. Any more information about how to repro would be helpful.

[euank]

Per above xref I was able to reproduce it and tracked it down (with @dgonyeo's help) to a bug in rkt. I think we can leave this issue open to track setting a version of rkt without said bug as the recommended version.

[yifan-gu]

@euank Are you able to test if the rkt 1.6 fixed this?