- Update and add more hooks (for new code that's about to be added).
- Address found issues.

New apartment, new internet connection. This time using fiber optics and
a _much_ faster connection. Change the port layout a bit. Change the
queueing a bit too.

The goal is to have all of the infrastructure that goes in building and
maintaining shore.co.il in a single repository. Why? Copy less code and
configuration around, maybe share some more code. Hopefully ease
maintenance. Maybe perform far-reaching changes in a single branch
instead of spreading it out across multiple repos (not really sure about
this).

Now that the Kodi host in running Debian Bullseye.

In case I need access to Nextcloud.

I think, not 100% sure. It's in use for a day now and seems to perform
better but opening new connections is still slow.

So they don't all fire at once. Less stress on each host or on the
entire infrastructure.

For accessing Nextcloud and testing a bunch of things.

- mnpw is now working (enough) so let's put it to use for all of the
  various Nagios-style checks.
- Add the monitoring plugins from monitoring-plugins.org/ for check_apt.
  Now I should get better alerts for pending updates. Regular updates
shuold be just a warning but security updates should be a critical
alert. Let's put it in production and see.
- Now with check_apt handling APT updates, write a small shell script
  for a Nagios-style check for firmware updates.
- And the last part of the updates, update flatpaks silently.
- Refactor of the general maintenance and btrfs-specific tasks.
- Tag the include_tasks tasks with always, so those tasks are always
  included so their tags are also evaluated.
- Remove the apt-transport-https package installation, it's a dummy
  package for apt since Debian Buster and Ubuntu Bionic.

This reverts commit fae72bee. Unexpected
consequences.

Makes the local resolver configuration more involved but I think it's
worth when the router is under heavy load.

- New DNS zone for nehes.co.
- Certificates for *.nehes.co.

command_warning is being deprected, callback_whitelist is being renamed to
callback_enabled.

Change the socket group so it can be accessed from a Toolbox container.

Service and timer to run needrestart and notify if a restart is needed.

It's already enabled by default in Bullseye.

In the upcoming Debian Bullseye, the docker.io package is new enough so
it functions properly with cgroup v2.

- Download firmware updates.
- Notify on pending updates.

I played around and caused a fuss with the secondary server. This puts
everything back in order.

Download package updates and notify if there are pending updates.

host01 is falling over running everything there. ns4 is secure enough
for the logs (volumes are encrypted, I don't think that there are
passwords or keys in the logs, just private information).

For configuring the wireless access point running OpenWrt.

The only reason it was a seperate role was because I thought it could be
useful for workstation. It is but it depends on a bunch of stuff from
the debian_server role. On a new host, it would fail to run before the
debian_server role has (it does because the debian_server role depended
on the toolbox role). When the need arises to use some of the
debian_server tasks in a workstation setting, it will need to be split
to a debian_common role.

It's already setup anyway by the playbook, so make it a hard dependency.