[BUG] Error while deploying kyverno on minikube

[shubham-asati]

Describe the bug
I have tried installing kyverno using installation.md on minikube .But on deployment pod gets into error phase.
Logs from pod

• init-container
  ERROR: logging before flag.Parse: I1227 14:20:57.721367 1 main.go:109] Using in-cluster configuration ERROR: logging before flag.Parse: I1227 14:20:57.722398 1 client.go:281] Starting registered resources sync: every 10000000000 seconds

• kyverno container
  standard_init_linux.go:211: exec user process caused "no such file or directory"

minikube version: v1.6.2
vm-driver: virtualbox
Kubernetes version : v1.17.0
Docker version : 19.03.5

To Reproduce
Steps to reproduce the behavior:

1. follow the minikube installation guide
2. run kubectl apply command as given in kyverno installation guide
3. Check pod status

Expected behavior
Successful installation

[shivdudhani]

init-container error is due to a missing flag.Parse() before using the arguments. Adding a init with flag arguments should fix it (nirmata@e6e5e3b)

[shivdudhani]

The other error with kyverno_container is due to an incorrect target in CI script.

The binary is linked to C libs and the vanilla alpine image does not contain the clib and this is due to default go flags defined during build.

The PR #582 should fix the above issue:

The specific make targets for initContainer & kyverno container disable the CGO flags while building the binary. Updating the CI script to use the specific make targets should resolve this issue.

The default make targets make build & initContainer still use the users go environment, but the CI will disable the flags for the build.

@shubham-asati I have tested the above fix with the following setup, do confirm if the above fixes the reported issues for your setup.

Testing Environment:
minikube version:v1.6.2
vm-driver: hyperkit
Kubernetes version : v1.17.0
docker version: 19.03.5

[realshuting]

I'm deploying the latest kyverno, seems like the pod falls in crashloopbackoff.
Images:

nirmata/kyverno                                          latest               9c36b66332da        2 hours ago         31MB
nirmata/kyvernopre                                       latest               3b4fc19d1c11        2 hours ago         27.4M

Logs:

• container kyverno-pre:

I1230 21:00:04.689992       1 main.go:115] Using in-cluster configuration
I1230 21:00:04.691756       1 client.go:281] Starting registered resources sync: every 10000000000 seconds

• container kyverno:

standard_init_linux.go:207: exec user process caused "no such file or directory"

[shivdudhani]

there was a typo in make target command, fixed in nirmata@d0ab83a

[shubham-asati]

The other error with kyverno_container is due to an incorrect target in CI script.

The binary is linked to C libs and the vanilla alpine image does not contain the clib and this is due to default go flags defined during build.

The PR #582 should fix the above issue:

The specific make targets for initContainer & kyverno container disable the CGO flags while building the binary. Updating the CI script to use the specific make targets should resolve this issue.

The default make targets make build & initContainer still use the users go environment, but the CI will disable the flags for the build.

@shubham-asati I have tested the above fix with the following setup, do confirm if the above fixes the reported issues for your setup.

Testing Environment:
minikube version:v1.6.2
vm-driver: hyperkit
Kubernetes version : v1.17.0
docker version: 19.03.5

@shivdudhani Thanks for the help . Both issues are fixed .