Docker hacks

Rename docker volume

Unfortunately Docker does not directly support volume rename operation. But same result can be achieved by copying data from the one volume to the other (assuming source and destination already exist):

$ docker run --rm -it -v <source>:/from -v <destination>:/to alpine ash -c "cd /from ; cp -av . /to"

Also through the SSH:

$ ssh <host> 'docker run --rm -v <source>:/from alpine ash -c "tar cf - -C /from ."' | docker run --rm -i -v <destination>:/to alpine ash -c "tar xpf - -C /to"

Late source can be removed.

Host user in Docker container

With bind mount, a directory on host machine is mounted into a container. And files in that folder created by process from the container will have owner name and group set to that process’s name and group (such as root).

Which is inconvenient since host user will not be able to change such files. To avoid this issue user with the UID and GID same to host user should exist inside container. Next commands in Dockerfile will do that:

ARG UID
ARG GID

RUN addgroup --gid $GID host-group
RUN adduser --disabled-password --gecos "" --uid $UID --gid $GID host-user

Corresponding values should be provided during build:

$ docker build --build-arg UID=$UID --build-arg GID=$GID .

User with the name host-user will exist inside container. And any files created by this user will look exactly the same like created by local user.

Alternative BASEIMAGE

Might be useful to build images for more then one CPU architecture with the same Dockerfile. Next commands will allow to specify base image during build:

ARG BASEIMAGE=debian:stretch-slim

FROM $BASEIMAGE

By default, build will use amd64 architecture. And to build image for armhf (Raspberry Pi):

$ docker build --build-arg BASEIMAGE=resin/rpi-raspbian:stretch .