Data directory

dataDir is a an option present in all the services, allowing users to specify the directory where a given service will store its data files. Essentially, it is to persist the state of the service even when you exit the process-compose window.

The dataDir of these services tend to take relative paths, which are usually relative to the project root. As such, when you run these services using nix run, their data files are created relative to whichever directory you are in. If you want these data files to always reside relative to the project directory, instead of using nix run consider wrapping the process-compose packages in script, via either mission-control module or a justfile. services-flake uses the latter.

Default data directory structure

Let’s say your project defines the following services:

{
    # Inside `perSystem.process-compose.<name>`
    services.postgres.pg.enable = true;
    services.redis.rd.enable = true;
}

The data directory structure will look like this:

|-- data
|   |-- pg
|   |-- rd

Gotchas

Unix-domain socket path is too long

Some services create unix domain socket files under the data directory. As the unix socket length is limited to about 100 chars, if your data directory is nested too deep, you will have to set dataDir option of the service to a shorter path as a workaround.