Uses php-fpm image and is served through a simple caddy server.
Automatically deploys by default and can be automatically deployed with
smtp e-mail sending and s3 primary object storage optionally if desired.
Utilizes some necessary hackery for container ordering and startup so
startup is relatively slow (takes around 2-5 minutes at least) but once
running should be stable and uninterrupted.
Implements health-checks for all involved containers.
Switch apache for php-fpm image
Added basic monica docker deployment.
A key which is required for monica is automatically generated on
deployment.
Optionally, geolocation and weather api can be enabled.
Mail-sending functionality is disabled by default but can be enabled
with any compatible smtp server set up.
Fixed access to postgres database not working for non-'miniflux'
usernames.
The database created is the same name as the username while the miniflux
container would still look for the standard database. This fixes it to
be able to take any username and corresponding database.
Contains default credentials but these should be overwritten with
group or host -destined variables.
Add default credentials
Hide real credential files from git
Added role of wallabag. Can take a subdomain alias to be presented on a
different uri than its stack name.
All services contain a healthcheck for docker swarm to be informed of
their status.
Copy whoami template to wallabag role
Change non-domain domain name to localhost
Fix correct handler notify naming
Allow setting different subdomain alias than name
Add meta information
Moved the ability to switch off caddy's api access to variables not
intended to be overwritten. It still can, but it is much more difficult.
Since every docker deployment relies on caddy's api to be accessible it
seems weird to present an easily reachable option to turn it off
globally.
Give more time to docker to correctly bring up caddy in the swarm.
Extended delay to retry to 10 seconds and gave it more retries as well,
so that roughly a minute will be gone before the play gives up.
If a domain has been declared for the host, the upstream file will proxy
at a subdomain named for the stack it proxies to,
otherwise it will do the same for the path.
Created bind volume for caddy which takes the json fragments which will
later be injected through the caddy api to dynamically set their
individual reverse proxy addresses.
This is not entirely necessary if the injection should be done as a
variable instead, but makes it easier to change routes and especially
more complicated ones (I'm thinking of gitea with ssh access e.g.).
Added caddy as swarm container. Uses templated Caddyfile which can be
set up for testing purposes for now.
Depends on docker and docker-swarm, and will keep its stuff where
docker-swarm sets itself up.
Variable `docker_stack_files_dir` can be used if other stack deployments
need to copy over configuration files or other extra files. Defaults to
`/stacks` but can be set to anything and other roles should take care to
respect this and put their files relative to it.
Variable `docker_swarm_public_network_name` can be used to set the name
of the network to be created that faces outwards (ingress network) and
to which all stacks will be connecting.
Added variable docker_swarm_advertise_addr which can be set to the
advertise address swarm should use.
Mostly introduced for vagrant-based test setups for now, which contain
multiple IP addresses and fail otherwise.
