Healthcheck needs review
I feel there's a race condition between the existence of the _koha DB and the koha container waiting. Also, start_period and start_interval are not being recognised on my swarm setup, but that might just be specifics of my setup.
Anyhow, having run.sh rely on the DB healthcheck might be problematic. I didn't manage to implement it to just do the SELECT 1;
as root. Something to keep looking at.
Having tweaked the healthchecks, and removed the db one, I've got a functional Koha in docker.