Docker image for syncing two docker volume contents to create a cache for slow networked mounts (osxfs/nfs).
In principle it is almost exactly the same as docker-bg-sync
which I have discovered after the fact, however, it uses two containers
(based on the same image) to ensure permissions (uid/gid) are always guaranteed
to be proper on both sides.
The image is published to docker hub as pink33n/docker-volume-sync.
See docker-compose.yml for a usage example.
Each container runs unison as it's mount owner (uid/gid) with sudo.
This is specified SYNC_UID and SYNC_GID env vars, however, the only
was to guarantee it can always access the files is to actually start
the container as the mount user instead of setting these variables.
The container can be ran in two modes:
- Server mode - starts a unison listener as the target user.
- Client mode - connects to the server and actually initiates the sync.
Start two containers for each volume, mounted at /dvs/data - one in
client mode - with SYNC_SERVER_HOST defined as the host address of
the second server mode container.
The client container should have the host (networked) volume mounted as
it performs the actual sync and is always the preferred source of truth
(prefer unison option).
Usually the containers should be started as the user (UID:GID) that owns the files in the volume. Alternatively you can specify the pair via env vars and the entrypoint will switch to this user (container user must be root for this). The second option might introduce problems with permissions / access and should be used only when you're sure the actual contents will be always owned by this uids.
For more configuration options see the default (overridable) env vars defined
in entrypoint.