Docker Sync

A script to synchronize docker images between hosts with emphasis on reducing the amount of data transfered. It is a pure python3 script, does not depend on the docker registry and has no dependencies on the remote server you are synchronizing with.

Typical use case

You are part of a team of 4 programmers working on a code base that runs on a docker image. Lets assume it is a Django image that has the following Dockerfile:

FROM python:3.4-slim

RUN apt-get update && apt-get install -y \
        gcc \
        mysql-client libmysqlclient-dev \
        postgresql-client libpq-dev \
        sqlite3 \
    --no-install-recommends && rm -rf /var/lib/apt/lists/*

ENV DJANGO_VERSION 1.8.6

RUN pip install mysqlclient psycopg2 django=="$DJANGO_VERSION"

Everything is working ok until you need a new package installed on the image. Now you have to add the package to the first RUN command, rebuild and push it to the registry for everyone to download.

These two RUN commands account for 240mb of the image and both layers are rebuilt, effectively making your whole team download these 240mb again from the registry every time a new dependency is added to the image.

With the command docker-sync user@somewebserver.com django:latest you would only transfer (with rsync) the compressed difference between the two images. In this case, if the package were for example gettext the whole update would only be around ~1mb and no need to push or pull from any docker registry.

Installation

curl -L https://github.com/dvddarias/docker-sync/raw/master/docker-sync > /usr/local/bin/docker-sync
chmod +x /usr/local/bin/docker-sync

Usage

Lets assume you want to synchronize your local machine docker images with the ones on your myamazingweb.com server. You would run:

>> python3 docker-sync user@myamazingweb.com

The output is something like:

Connecting to user@myamazingweb.com.
Listing user@myamazingweb.com images...............................DONE
Listing local images........................................DONE
--> Report:
current: ['squid:latest', 'ubuntu:latest', 'debian:8.0']
need update: ['redis:latest']
new: ['mongodb:latest', 'gogs:latest']

Nothing was synchronized yet, this is just to see the status of all the images:

Current means the images have the exact same ID.
Need Update means that the two images have the same name:tag but different ID.
New means you dont have a local image with the listed tag.

To sync the redis image run

>> python3 docker-sync user@myamazingweb.com redis:latest

Please run the script with --help for further details.

How does it work

When synchronizing two images it connects via ssh to the remote server and:

Dumps both images (local & remote) to the hard drive (docker save).
Runs rsync over the ssh connection to synchronize the content of both files
Loads the new image (docker load)
Removes the image dumps on both computers.

Note: When synchronizing an image on the New list the script will try to look for the highest common parent on your local machine to run the rsync command, so it pays off a lot to use images with the same base image. If no common parent is found the whole image is compressed and transfered.

Dependencies & Configuration

You need python > 3.4 on your local machine and both the user running the script and the user on the remote host most have permissions to run docker commands. It uses ssh to connect to the host so you should also have the the appropriate permissions. There are no dependencies on the remote host other than: bash, rsync & ssh that are already installed on most linux distributions.