Fixing and creating directories about s6-overlay

[glerchundi]

Can you explain a little bit more? Or give a full example in order to assess if it's worth it or not?

[analogic]

Better example: service wont run without log directory - for ex. nginx. I mount volume /var/log - for first time its completely empty and directory /var/log/nginx is missing, nginx wont run until i create it with proper rights.

1. it would be nice to have some flag in fix-attrs, which will automaticaly run mkdir -p /var/log/nginx if directory not exists
2. or/and have ability to run a script BEFORE fix-attrs which will run mkdir -p /var/log/nginx

[analogic]

I would post a patch but execlineb is not my favorite language yet :)

[glerchundi]

@analogic sorry for the delay, right now, i don't have much time to invest on the overlay. Have you thought on making them in your container before building it up?

Something like this (take a look to /var/log/{nginx-access-logs,nginx-error-logs}): https://github.com/just-containers/nginx-loadbalancer

WDYT?

[glerchundi]

@analogic any news on this?

[analogic]

Sorry for delay - nginx was only for example - there are other services which needs precreate directories with proper rights. I am author of https://hub.docker.com/r/analogic/poste.io/ and handling directories which are exposed to users is kind of complex task. :) I will write patch when i will have some time...

[glerchundi]

My nginx references were an example too. I pre-create folders within the rootfs before building the image itself, it's more declarative and seems cleaner, at least for me. Give it a try and tell me what you think and if this doesn't fit explain your use-case a little bit more. Image sources would help of course.

[analogic]

Technicaly my image is blackbox with only datadir/logs exposed via volumes - rootfs is not problem at all. Mainly i am solving first container run (precreating needed directories), fixing user errors (deletion some of a logs dir) or updates in datadir structure... its no big deal, but it would be nice to switch all this handling from custom script to simple fix-attrs

[glerchundi]

What about something like this (this is the way it was done in nginx-loadbalancer):

mysql-data-image/Dockerfile:

FROM busybox
VOLUME /var/lib/mysql
COPY rootfs /
CMD ["true"]

mysql-data-image/rootfs/var/lib/mysql/test-file:

buh from a data container

Then:

1. docker build -t mysql-data-image .
2. docker run --name mysql-data mysql-data-image
3. docker run --rm --volumes-from mysql-data quay.io/justcontainers/mariadb:v0.4.0 cat /var/lib/mysql/test-file

Among other things this is printed to the stdout:

[...]
buh from a data container
cat exited 0
[...]

[glerchundi]

@analogic please close this whenever you can.

[analogic]

sorry, i am bit busy right now... ill do patch later