I began using LinuxServer images quite recently and I can say that I am in love!
The simplicity and the fact that it simply works are big wins for me.
Great job and thank you so much! You deserve all the kudos you can get!
However, as I move from messing around with docker-compose serving a semi-working media server and IRC client on my workstation, to a rpi powered docker swarm serving a fully-working high-availability media server (I have very high demanding clients: my SO), I began to question the usage of bind mounts over named volumes (actually, I have been troubled my this since day zero )
Throughout the docker documentation we are always encouraged to use named volumes over bind mounts. And I understand the concept, if docker is capable of handling my storage needs why should I care to maintain the data myself outside of docker?
Slap a driver into the container (like GlusterFS) and you have a replicated volume, great!
Even at work, I’ve had contact with subcontracted docker (supposedly) experts and they always advocate de usage of named volumes over bind mounts (but the also advocate the backing up of data with the services running for “zero downtime backup” so…).
However, I can also understand the need/will to use bind mounts here at LinuxServer: not only data but also configurations live in the
/config folder (editing a configuration file inside a named volume is considerably more “dificult” than editing them in a bind mount).
But, it does feel … unnatural to use bind mounts over named volumes.
I mean, there are some cases where I definitely want to use bind mounts, I have no space in all me nodes to replicate my media library, for example.
All that taken into account I have, in fact, three questions:
Why does LinuxServer advocates the usage of bind mounts over named volumes?
Is there anybody using named volumes over bind mounts out there?
If you are, how the hell do you “inject” your configurations into the volumes?
Many thanks in advance.