search

denverdino / lxcfs-admission-webhook

Apache License 2.0
159 stars 81 forks source link

after lxcfs container crashed it could not be restart again. #4

Closed g0194776 closed 4 years ago

g0194776 commented 4 years ago

lxcfs container logs:

container_linux.go:247: starting container process caused "process_linux.go:359: container init caused \"rootfs_linux.go:53: mounting \\\"/var/lib/lxcfs/proc/uptime\\\" to rootfs \\\"/var/lib/docker/devicemapper/mnt/80cc788550e01d8690970b31e471b16d0c3093e35edba07d427ca305c3fd3e7c/rootfs\\\" at \\\"/var/lib/docker/devicemapper/mnt/80cc788550e01d8690970b31e471b16d0c3093e35edba07d427ca305c3fd3e7c/rootfs/proc/uptime\\\" caused \\\"not a directory\\\"\""
[root@192 proc]# ll
total 0
drwxr-xr-x. 2 root root 6 Dec 22 21:23 cpuinfo
drwxr-xr-x. 2 root root 6 Dec 22 21:23 diskstats
drwxr-xr-x. 2 root root 6 Dec 22 21:23 meminfo
drwxr-xr-x. 2 root root 6 Dec 22 21:23 stat
drwxr-xr-x. 2 root root 6 Dec 22 21:23 swaps
drwxr-xr-x. 2 root root 6 Dec 22 21:23 uptime

On the problem machine, the path of /var/lib/lxcfs/proc/cpuinfo was changed to directory type which originally is a file, that's strange.

@denverdino

g0194776 commented 4 years ago

Due to mount points in the host, the lxcfs container COULD NOT restart successfully again. New created lxcfs container will stuck at Created status until you manually unmount the points from host, That's too bad.

We've tried, and use RPM to install lxcfs instead of running containers by DaemonSet in the Kubernetes.