CODEWITHSUNDEEP
linuxdocker
Mr Pang
Mr Pang

Reputation: 1181

Can I run host commands with docker container environment?

Many docker images have no base commands such as ifconfig, ip, ps, etc. So I would like to find a way to run host commands in docker container environment so that I don't need to copy/pack every commands to docker image.

I know a command ip netns exec [ns] ifconfig, which can run the command ifconfig in a new network namespace. Is there similar way for docker container?

Upvotes: 1

Views: 1273

Answers (1)

Danila Kiver
Danila Kiver

Reputation: 3758

This question misses the fact that the kernel does not have a notion of a "container" - from its standpoint, the container is a set of different abstractions (different namespaces, cgroups, etc.). Having enough rights on the host, you can manage these abstractions independently (e.g. enter any single namespace of the container, or a couple of them). That said, the full answer to the question depends on what "execute the host's binary in the container" means (do you want to execute this binary in all namespaces of the container, or only in significant ones, which influence the result you want to obtain?).

Technically, you can execute arbitrary host's binary inside the container's namespaces (except for mnt namespace) without copying or mounting this binary into the container's filesystem. Depending on what you are trying to achieve, it may or may not solve your problem.

For example, doing some ip stuff after entering only net namespace of the container will probably give you the expected result:

$ sudo nsenter -t $CONTAINERIZED_PROCESS_PID -n ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
33: eth0@if34: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default 
    link/ether 02:42:ac:11:00:02 brd ff:ff:ff:ff:ff:ff link-netnsid 0

The command from the host gets executed in the container's network namespace and shows its network interfaces correctly.

Or you can get correct container's hostname executing hostname binary inside container's UTS namespace:

$ sudo nsenter -t $CONTAINERIZED_PROCESS_PID -u hostname
d65269ecf908

However, doing ps after entering pid namespace of the container will not give you the list of the container's processes, but rather the list of all processes you are able to see from the host, because ps takes the information from /proc, so to see the correct set of processes you have to enter the container's mnt namespace (which in turn will make you unable to run the host binaries, because you will no longer see them).

That said, it really depends on what do you want to execute and what do you want to get as a result.

Upvotes: 2

Related Questions