问题
I built a customised Docker image of airflow following this document: "https://github.com/puckel/docker-airflow". Built and run in my local VM. Everything was successful and airflow was up. Pushed the image to ACR (Azure container registry) and launched it in aks via stable helm chart. Referred this link "https://github.com/helm/charts/tree/master/stable/airflow". Now suddenly in kubernetes the pods are not up and fails out with the below error.
Error: failed to start container "airflow-scheduler": Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused "exec: \"/usr/bin/dumb-init\": stat /usr/bin/dumb-init: no such file or directory": unknown
Back-off restarting failed container
The most weird part of this error was I was able to do this till the previous day without any issues. Suddenly today this error pops in AKS alone. It runs fine in my local VM too successfully. Have anyone experienced this before?
P.S. for those who don't know the stable chart has an option to give our custom image URL in the values.yaml file. I'm also using the same method to deploy my image.
回答1:
Found out what issue was. stable/airflow chart had made a massive update to 7.0.1 and one of those changes involves not supporting puckel-docker image anymore and will support only apache's official docker image going forward. Can look at this issue to get an insight https://github.com/helm/charts/issues/22483
来源：https://stackoverflow.com/questions/61955393/weird-error-in-kubernetes-starting-container-process-caused-exec-usr-bin