12:04 a.m.
On lip 28, Robin Schneider wrote:
Also, this common "interface" would probably only work for
some roles. For the
databases MariaDB and PostgreSQL, it would not as they have different concepts.
These databases are not mutually exclusive - both databases can be present on
the same host, and only application role decides which one to use. There's no
need for a wrapper role here, 'mariadb' and 'postgresql' roles in the
playbook
enabled/disabled conditionally is probably the best we'll gonna get.
Don’t be afraid of SaltStack itself :) SaltStack can make use of
Ansible at
various levels (module level, playbook level). The other way around seems to
also have some integrations.
True, but base installation of 'salt-minion' on Debian Buster wants to take
38 MB with all the various dependencies. In comparsion, mosquitto takes 1 MB.
Looking around a bit I stumbpled upon loopabull project[1] which can execute
Ansible playbooks based on events from various message buses. It might be an
interesting side project to explore, if we find some uses for it later on.
[1]: https://github.com/maxamillion/loopabull
But I would rather ask if there are any technical reasons against
things like
chrony. Otherwise you have the issue that people doing the survey might not be
up-to-date/universally informed and just vote for what they currently use. We
should still push for good solutions rather than legacy once.
I didn't look at chrony in detail, probably because most of the time I work
inside LXC containers these days. When the 'ntp' role is split, we can change
to chrony as the default - I'll probably remove 'ntp' role entirely from the
common playbook so that it's easier to install NTP service only on hardware
and VMs where it makes sense. The timezone configuration has been moved to the
'tzdata' role for this purpose.
-- Maciej