Mutagen-based file synchronization

Estimated reading time: 6 minutes

Docker Desktop for Mac on Edge has a new file sharing feature which performs a continuous two-way sync of files between the host and containers using Mutagen. This feature is ideal for app development where:

  • the source code tree is quite large
  • the source is edited on Mac
  • the source is compiled and run interactively inside Linux containers.

This page contains an example to show how the Mutagen feature should be used to sync files between the host and containers. It also gives some best practices to maximize performance.

Important

After completing the synchronizing process, you must delete and re-create any containers which will make use of the synchronized directories.

A simple React app

The following example bootstraps a simple React app with npx and configures Docker Desktop to sync the source code between the host and a development container.

First, create a directory which will contain the app:

$ mkdir ~/workspace/my-app

Next, enable the two-way file sync feature in Docker Desktop:

  1. From the Docker Desktop menu, click Preferences > Resources > File sharing.

  2. Type the new directory name in the bottom of the list and press the enter key.

  3. Click on the toggle next to the directory name.

  4. Click Apply & Restart for the changes to take effect.

When Docker Desktop has restarted, the File sharing page looks like this:

Caching with mutagen is "Ready"

Run the following command to start a container and bootstrap the app with npx:

$ docker run -it -v ~/workspace/my-app:/my-app -w /my-app -p 3000:3000 node:lts bash
root@95441305251a:/my-app# npx create-react-app app
root@95441305251a:/my-app# cd app
root@95441305251a:/my-app# npm start

Once the development webserver has started, open https://localhost:3000/ in your browser and observe the app is running.

Return to the File sharing page in the UI and observe the status of the synchronize toggle located next to the directory name. The status will be updated as file changes are detected and then synchronized between the host and the containers.

Wait until the text says Ready and then open the source code in your IDE on the host. Edit the file src/App.js, save the changes and observe the change on the webserver.

As you edit code on the host, the changes are detected and transferred to the container for testing. Changes inside the container (for example, the creation of build artifacts) are detected and transferred back to the host.

Adding synchronized directories through the CLI

As of Docker Desktop Edge 2.3.2.0 the :delegated flag on a shared volume will automatically enable synchronization. This is particularly useful if you want to share the synchronization setting with other developers in your team, through your source code.

Scripting

Many projects have scripts which modify files in a container and then read them back on the host, or the other way around. To support these, Docker Desktop automatically flushes file changes around docker run calls. Therefore, the following will work as expected:

docker run -v ~/foo:/foo:delegated touch /foo/new-file-in-container
stat ~/foo/new-file-in-container

and

touch ~/foo/new-file-on-host
docker run -v ~/foo:/foo:delegated stat /foo/new-file-on-host

Note that changes are not flushed around docker exec calls, so this will not work as expected:

docker exec existing-container touch /foo/new-file-from-existing-container
stat ~/foo/new-file-from-existing-container # file will not be found

Bypassing a two-way sync for a volume

By default, if a directory has two-way sync enabled, then shared volumes will use the sync. However, if you want to avoid using a two-way file synchronization, use :consistent: this will bypass both file caching and the two-way sync. This is backwards compatible with previous Docker versions.

Avoiding synchronizing a subdirectory from a container to the host

Although two-way file sync is suitable for many types of files, sometimes containers can generate lots of data which doesn’t require copying to the host, for example, debug logs.

If your project has a subdirectory that doesn’t need to be continuously copied back to the host, then use a named docker volume to bypass the sync.

First create a volume using:

$ docker volume create donotsyncme
donotsyncme

Use the volume for the subdirectory you want to avoid syncing:

$ docker run -it -v ~/workspace/my-app:/my-app -v donotsyncme:/my-app/dontsyncme -w /my-app -p 3000:3000 node:lts bash

Docker Desktop will sync all changes written by the app to /my-app to the host, except changes written to /my-app/dontsyncme which will be written to the named volume instead.

Avoiding synchronizing a subdirectory in both directions

Although the example in the previous section successfully prevents changes written in the container from being copied back to the host, changes written on the host will still be written to the container (although hidden by the docker volume).

To completely avoid synchronization in both directions, create a global mutagen config file. For example, create a file ~/.mutagen.yml containing:

sync:
  defaults:
    ignore:
      paths:
        - /path1/to/ignore
        - /path2/to/ignore
        - ...
        - /pathN/to/ignore

Best practices

To achieve maximum performance when you enable two-way file sync:

  • Avoid wasting disk space and CPU by minimising the size of the synchronized directories. For example, synchronize a project directory like ~/my-app, but never sync a large directory like /Users or /Volumes.

    Remember that the files will be copied inside the container and therefore must fit within the Docker.qcow2 or Docker.raw file.

  • For every volume you want to sync in docker run -v or in docker-compose.yml, ensure either the directory itself or a parent/grandparent/... directory is listed in Preferences > Resources > File sharing.

    Note in particular that if only a child directory is listed on the File sharing page, then the whole docker run -v may bypass the two-way sync and be slower.

  • Avoid changing the same files from both the host and containers. If changes are detected on both sides, the host takes precedence and the changes in the containers will be discarded.

  • After completing the caching process, you must delete and re-create any containers which will make use of the synchronized directories.

Feedback

Your feedback is very important to us. Let us know your feedback by creating an issue in the Docker Desktop for Mac GitHub repository with the Mutagen label.

mac, mutagen, volumes, file sync