localkube leaving 'wc' zombies

[ankon]

Is this a BUG REPORT or FEATURE REQUEST? (choose one):

BUG REPORT.

Minikube version (use minikube version):
minikube version: v0.14.0

Environment:

• OS (e.g. from /etc/os-release): Fedora 25 x86_64
• VM Driver (e.g. cat ~/.minikube/machines/minikube/config.json | grep DriverName): kvm
• Docker version (e.g. docker -v): Docker version 1.12.5, build 6009905/1.12.5

Noticed the issue both with minikube v0.14.0 and v0.13.1, cannot remember whether I saw it with earlier versions.

What happened:

I'm looking inside the VM with top, and I see a lot of wc processes in Z state, under the localkube process. Watching it for a while: There is a du process that quickly goes away, and then a wc, which sometimes but not always doesn't seem to get reaped. Eventually the VM can have hundreds of these zombies.

What you expected to happen:

Zombie-free environment :)

How to reproduce it (as minimally and precisely as possible):

For me it just happens when running the VM with some pods in it, nothing "special" is done.

Anything else do we need to know:

Guess: It's somewhere in vendor/github.com/google/cadvisor/fs/fs.go, which seems to talk about du and wc processes. My go-foo isn't anywhere close to knowing how to debug it further.

[r2d4]

This has already been fixed in the new ISO
http://storage.googleapis.com/minikube/iso/buildroot/minikube-v1.0.1.iso

However, we haven't backported a fix to the b2d ISO yet.

ref #923

When find exits with an error, the wait for the wc process never happens and wc becomes a zombie. On systems without GNU find, these zombies can accrue quickly.

This PR looks like it fixes the problem upstream google/cadvisor#1558

[ankon]

Thanks for the reply.

Quickly tested it with minikube start --vm-driver virtualbox --iso-url https://storage.googleapis.com/minikube/iso/minikube-v1.0.1.iso --kubernetes-version v1.4.5 --memory 4096, looks good!

[dlorenc]

We're not going to backport this to the old iso, this will be fixed in the new ISO in our next release.