I discovered the existance of s6 overlay and cont-init through the linuxserver images. I’m curious for the motivation for it, since each of the containers (at least the ones I’ve looked at) only run a single process anyway.
From what I can see:
- Simple instrumentation of multiple processes
- More work to override command (if I for example want to add command-line flags to the process I essentially have to override the init script by either building my own image or shriek mounting the config as a volume).
Possibly unrelated, but I’m having some issues with sonarr and radarr not stopping properly when told to - since the owning process has PID 1, this shouldn’t be related, but since I haven’t found the root cause yet I can’t fully discount it yet - basically the s6 approach adds another (albeit thin) layer of complexity that can’t be discounted with this class of issues.
So I see more cos than pros, which makes me curious; how does the lsio team think about it and what was the reason for sticking with this approach?
(This was intended for the lsio category but being a new user I don’t have permission to post there)