When I create a new docker container like with
docker run -it -m 560m --cpuset-cpus=1,2 ubuntu sleep 120
and check its namespaces, I can see that new namespaces have been created (example for pid 7047).
root@dude2:~# ls /proc/7047/ns -la
total 0
dr-x--x--x 2 root root 0 Jul 7 12:17 .
dr-xr-xr-x 9 root root 0 Jul 7 12:16 ..
lrwxrwxrwx 1 root root 0 Jul 7 12:17 ipc -> ipc:[4026532465]
lrwxrwxrwx 1 root root 0 Jul 7 12:17 mnt -> mnt:[4026532463]
lrwxrwxrwx 1 root root 0 Jul 7 12:17 net -> net:[4026532299]
lrwxrwxrwx 1 root root 0 Jul 7 12:17 pid -> pid:[4026532466]
lrwxrwxrwx 1 root root 0 Jul 7 12:17 user -> user:[4026531837]
lrwxrwxrwx 1 root root 0 Jul 7 12:17 uts -> uts:[4026532464]
root@dude2:~# ls /proc/self/ns -la
When I check with ip netns list
I cannot see the new net namespace.
dude@dude2:~/docker/testroot$ ip netns list
dude@dude2:~/docker/testroot$
Any idea why?
That's because docker is not creating the reqired symlink:
# (as root)
pid=$(docker inspect -f '{{.State.Pid}}' ${container_id})
mkdir -p /var/run/netns/
ln -sfT /proc/$pid/ns/net /var/run/netns/$container_id
Then, the container's netns namespace can be examined with ip netns ${container_id}
, e.g.:
# e.g. show stats about eth0 inside the container
ip netns exec "${container_id}" ip -s link show eth0