Filesystem Sync

Outrigger uses Unison to provide high-performance, bi-directional file synchronization between your local operating system and Docker containers. This is not the only means of sharing code or data with application containers, but it is the recommended approach for sharing files that must also be efficiently available to developers and their local tools (e.g., IDEs).

With Unison, all the files from the root of your project are synced via a dedicated container and exposed as a named Docker Volume to the rest of your application. This volume is used as the file "source" for a volume mount to allow nearly native filesystem performance and features.

For more on why Outrigger introduced Unison, jump down to Why Sync Instead of NFS?

Setting Up Sync for Your Project's Containers

To setup unison sync for your containers you will need to reference an external volume in your docker-compose.yml and use that external volume in the mount specification for any services that need to reference the same set of files.

Since volumes are global within your Docker host, the name of this external volume should be namespaced to your project. The convention is to use projectname-sync as the volume name. Using the same volume will ensure all containers stay in sync. This includes specifying that volume for your build container, if you are using one (typically defined in build.yml).

Add the volume to your docker-compose file

Note that volume definitions are available as of docker-compose schema v2. volumes are a root-level property in the docker-compose schema and should not be nested below services.

volumes:
  projectname-sync:
    external: true

Note about volume name

If you optionally specified a volume name to the project sync:start command or configured a name in your Outrigger Project Configuration file the name in this volumes section must match the one specified. The reason why we don't exclusively grab the volume name from the compose file is that this volume may be managed outside of Outrigger by another tool. We also want to support named volumes that don't precisely follow the *-sync convention described here.

Use the external volume as your local mount point

services:
  build:
    image: outrigger/build:php71
    volumes:
      - projectname-sync:/var/www

Managing the File Sync Process

The file synchronization is managed by the rig cli via rig project sync:start and rig project sync:stop.

Configuring the Sync Volume Name

The name of the sync volume can be set in a variety of ways. It is determined using the following precedence:

  1. Specified in the sync -> volume-name property of your Outrigger Project Configuration
  2. In your docker-compose.yml file, specified as an external volume with a name of the pattern *-sync
  3. If not specified anywhere else, it will use the name of the current project folder with -sync appended

We recommend using approach #1 or #2 to ensure consistency across all environments. Approach #2 is most often used as it requires the most minimal configuration additions. If you use approach #1, make sure a volume of the same name is also defined as part of your docker or docker-compose configuration. Approach #1 is useful when the name of the volume can not follow the *-sync naming pattern for some reason.

Configuring ignores

When a need exists to restrict which files are synced using the ignores section. An example of limiting the synced files can be found in the sample in the Outrigger Project Configuration documentation.

Start it up

From your project root run the command rig project sync:start (there is also an alias rig project sync) to get going. The initial sync can take a few seconds depending on the size of your project folder. You will see a progress indicator that will let you know when the initial sync is finished and things are ready to use.

Workflow Change for Unison-using Projects

Projects that use this file syncing approach will require execution of this command as a new step before any other docker commands can be run.

Cleaning up

Since the project sync:start command starts another container to manage the volume syncing, we have a command project sync:stop that will clean up that running container for you. It discovers the volume / container name the same way project sync:start does.

When you are done for the day, or for that project, run rig project sync:stop from the project root to clean up any running sync containers for that project.

How File Sync works in Linux Environments

Linux environments have native filesystem performance without anything as complex as file sync, because they do not need to use a virtualization layer to have access to the Linux Kernel's container support.

For this reason, any environment running on Linux will have local bind volumes created automatically where a Unison/sync volume would be used. This allows for the use of named volumes, but avoids using Unison where local filesystems have all the functionality needed.

Setting up file sync manually using Override Files

See Understanding multiple Compose files to understand how Docker Compose files work.

Example docker-compose.yml

services:
  www:
    image: outrigger/apache-php:php71
    volumes:
      - .:/var/www

Example docker-compose.override.yml

services:
  www:
    volumes:
      - projectname-sync:/var/www
volumes:
  projectname-sync:
    external: true

Why Sync Instead of NFS?

Outrigger provides easy, NFS-based filesystem mounts to share code and files with your project containers. However, it is slow when writing, reading, or scanning thousands of files when compared to native filesystem performance.

Some modern tool kits and package managers favor large numbers of small files and libraries so maintaining high performance of a build becomes challenging when your code base is on a volume mounted NFS share.

NFS also has the downside of not propagating filesystem modification events. This can be problematic if you want to run a process in your container to rebuild or update compiled project assets when modification notifications are triggered due to local file editing in your IDE.

Given the current state of the art of Virtual Machines and filesystems, these types of operations are better supported by bi-directional file syncing (which includes filesystem notifications).