If docker starts before the filesystem gets mounted, you could be seeing issues with the docker engine trying to write to the parent filesystem. You can restart the docker daemon to rule this out (systemctl restart docker in systemd base environments).

If restarting the daemon helps, then you can add a dependency between the docker engine and the external filesystem mounts. In systemd, that involves an After= clause in the unit file. E.g. you could create a /etc/systemd/system/docker.service.d/override.conf file containing:

[Unit]
After=nfs-client.target

(Note that I'm not sure that nfs-client.target is the correct unit file for your filesystem, you'll want to check where it gets mounted.)

Another issue I've seen people encounter recently is Snap based docker installs, which run docker inside of another container technology, which would prevent access to paths not explicitly configured in the Snap.