仓库源文，站点原文

title: "3 Ways to Access Host System in a Docker Container" date: 2022-08-08T20:04:00+08:00 categories: [Kubernetes] tags: [docker]

aliases: [/posts/3_ways_to_access_host_system_in_container/]

When we are using Docker, we always access the host system by mounting the host folders as a volume. But sometimes we can not do that due to deployment reasons or security limits.

There are three workaround ways to access the host system.

1. Mount docker.socks into container

docker.socks is a Unix socket that Docker Engine API listens on. You can mount /var/run/docker.socks file to your container and call Docker Engine API through this socket.

For instance, If I want to get docker information by calling that API:

curl -s --unix-socket /var/run/docker.sock http://localhost/info

You can refer to Docker Engine API to find more useful details.

2. Using PID mode host

By default, Docker uses the Linux PID namespace to isolate containers' filesystem view. It means if two processes have the same PID, they will share the filesystem permission.

Docker has a startup parameter --pid=host to change the PID namespace to its host process's namespace. You can use this parameter to make your docker container have the same privilege as the host process.

When using --pid=host, you can list host system's processes by ps -ef.

3. By docker --privileged

It is the last way you can access the host system, but it is not recommended for most cases.