I have a container that wants to communitate with Kanto CM over gRPC. That's why my container is created with moun point like
--mp="/run/container-management/container-management.sock:/run/container-management/container-management.sock".
It is all fine, container is created and started successfully through the kanto-cm CLI.
However, the problem occurs when Kanto container-management service is restarted. Then my container remains stopped and not starting automatically.
The following message is seen in the kanto-cm log file.
level=error msg="[container-management][ctrd_client.go:255][pkg:github][func:com/eclipse-kanto/container-management/containerm/ctr] error creating task for container ID = 1eb42aec-8bb9-4da7-833e-fa3030a245a5 \n\t Error: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error mounting \"/run/container-management/container-management.sock\" to rootfs at \"/run/container-management/container-management.sock\": stat /run/container-management/container-management.sock: no such file or directory: unknown "
I guess the loading and restarting of previously available containers is performed before the starting of Kanto-CM gPRC server, that's why the container-management.sock file does not yet exist and mounting fails.
But could it make a container restart attempt later? What about container's restart policy, it seems to be ignored in this case...
A workaround is to use a mount point only to --mp="/run/container-management:/run/container-management".
I have a container that wants to communitate with Kanto CM over gRPC. That's why my container is created with moun point like
--mp="/run/container-management/container-management.sock:/run/container-management/container-management.sock".It is all fine, container is created and started successfully through the kanto-cm CLI.
However, the problem occurs when Kanto container-management service is restarted. Then my container remains stopped and not starting automatically.
The following message is seen in the kanto-cm log file.
level=error msg="[container-management][ctrd_client.go:255][pkg:github][func:com/eclipse-kanto/container-management/containerm/ctr] error creating task for container ID = 1eb42aec-8bb9-4da7-833e-fa3030a245a5 \n\t Error: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error mounting \"/run/container-management/container-management.sock\" to rootfs at \"/run/container-management/container-management.sock\": stat /run/container-management/container-management.sock: no such file or directory: unknown "I guess the loading and restarting of previously available containers is performed before the starting of Kanto-CM gPRC server, that's why the
container-management.sockfile does not yet exist and mounting fails.But could it make a container restart attempt later? What about container's restart policy, it seems to be ignored in this case...
A workaround is to use a mount point only to
--mp="/run/container-management:/run/container-management".