add support for hot reloading the config.yaml

[arnesetzer]

As discussed in #99 a try to enable hot reloading for config (and therefore datasets) without rebuilding the docker image again.

What changed:
A watchdog (/app/watcher.py) is etablish via supervisor. On a config.yaml add or change (CAVE: Due to docker limitations it won't happen if the inode changes, see: https://medium.com/@jonsbun/why-need-to-be-careful-when-mounting-single-files-into-a-docker-container-4f929340834) the watchdog is triggerd an kills the uwsgi and flushes memcached.

[arnesetzer]

@ajnisbet Any feedback? 🙈

[ajnisbet]

Thanks for this Arne!

It's no my radar! I've been on holiday the last month and there's a few things I'll need to tweak before merging, here's a quick review (for my later reference):

• Use subprocess instead of os.system
• Stop and start uwsgi with supervisord instead of killing it
• Add warm cache command
• Watch both config.yaml and example-config.yaml

[arnesetzer]

Hi, thanks for the feedback.

* [ ]  Use `subprocess` instead of `os.system`

Done, now using pymemcache to avoid any system manipulation.

* [ ]  Stop and start `uwsgi` with `supervisord` instead of killing it

I'm a little bit puzzled how to achieve this. I played around with the supervisorctl but this requires some serious changes in the supervisord configuration (

* [ ]  Add warm cache command

Already done 😄

* [ ]  Watch both `config.yaml` and `example-config.yaml`

Also done now.

[ajnisbet]

Thanks for this work @arnesetzer !

Here's the final working code if you're interested: d1fc1a7