Harden Runner breaks `medyagh/setup-minikube`

[tippmar-nr]

I'm not sure where the problem lies, but there's some odd interaction between the harden-runner action and medyagh/setup-minikube.

If I add harden-runner with egress-policy: audit, the setup-minikube step doesn't fully install and configure Minikube.

A sample workflow run showing the with and without behavior is available here: https://github.com/tippmar-nr/testing/actions/runs/9210727454 -- with harden-runner, the job runs for several minutes and emits a lot of odd messages, before finally failing because Minikube isn't running correctly. Without harden-runner, the job runs for just a couple of minutes and behaves exactly as it should.

Would appreciate any guidance or troubleshooting tips. Thanks!

[varunsh-coder]

@tippmar-nr thanks for reporting this and using harden-runner! I will look at the workflow run and investigate.

[varunsh-coder]

@tippmar-nr I noticed the https://github.com/tippmar-nr/testing repo no longer exists. Can you please share the workflows you used? Thanks!

[tippmar-nr]

@varunsh-coder Sorry about that - my testing repo was set to private instead of public. You should be able to access it now. Thanks for taking a look!

[varunsh-coder]

This is because harden-runner is unable to resolve domain: host.docker.internal
https://github.com/tippmar-nr/testing/actions/runs/9210727454/job/25338401257#step:12:56

This will need a fix to resolve domains that point to internal IP addresses.

[tippmar-nr]

Thanks for the update. I probably should have looked through the workflow output a bit more to identify that specific issue.

[varunsh-coder]

This issue only affects the Community tier. The Enterprise tier relies on eBPF and therefore does not encounter this issue.
We are still investigating how to best solve this for the community tier.