Init script fails when running on Kubernetes

[sauldhernandez]

When using latest version, startup fails with the following error:

2017-01-19T20:39:48.090681185Z Error: can't get my container ID !

I noticed that the error is generated by the startup script, since it cannot find docker on /proc/self/cgroup. Kubernetes mounts all cgroups on the root. So instead of what you would get starting a container with docker itself:

10:net_cls:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
9:freezer:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
8:memory:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
7:cpuset:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
6:perf_event:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
5:cpu,cpuacct:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
4:blkio:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
3:devices:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
2:pids:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508
1:name=systemd:/docker/25ffc6544832215a020964755b584753464bdafa49357b0552a9c51673ee8508

You get this in Kubernetes:

10:net_cls:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
9:freezer:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
8:memory:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
7:cpuset:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
6:perf_event:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
5:cpu,cpuacct:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
4:blkio:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
3:devices:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
2:pids:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31
1:name=systemd:/53c173f34a43de2ee8197f5efe5aca38747389db6f2c7c2d1d2cb2a93102ee31

Changing the lookup pattern so it looks for the container ID after the last / should probably fix the problem.

[JrCs]

For kubernetes it's better to use traefik

[myoung34]

#300 can help to resolve this if people still wish to use this container on k8s/ecs/etc

[buchdag]

This should no longer be relevant after #491 #499 and #516