You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think the role is maybe trying to do too many things. It can bootstrap a VPS, install Decidim but also upgrade an existing Decidim.
Most tech shops have a way to install a VPS and configure DNS the way they like, so I would potentially drop that functionality.
Ansible has historically been bad for dealing with ongoing app maintenance. It shines for server provisioning however. I would drop the "upgrade Decidim" path and remove the need for "complete_install: true/false`. The role then specialises in "install from scratch".
The Redis package is installed via Snap and runs into issues when it tries to restart via systemctl. I would avoid installing via Snap and stick to the system packages.
Some stuff seems "left over" in the role, e.g. vars for matomo. It would be good to look at what vars are still used and what is essential. Perhaps also the sentry stuff is unused?
For a role with this level of complexity (several sub-roles, vars which change the execution flow), I would recommend setting up routine Molecule testing so the role can be guaranteed to be working.
You have several things to test already: Ansible version, Ruby version, Node version and Ubuntu version. And in between all the system dependency versions, e.g. Passenger. Without regular testing, this role can become unreliable.
The text was updated successfully, but these errors were encountered:
I think the role is maybe trying to do too many things. It can bootstrap a VPS, install Decidim but also upgrade an existing Decidim.
Most tech shops have a way to install a VPS and configure DNS the way they like, so I would potentially drop that functionality.
Ansible has historically been bad for dealing with ongoing app maintenance. It shines for server provisioning however. I would drop the "upgrade Decidim" path and remove the need for "complete_install: true/false`. The role then specialises in "install from scratch".
The Redis package is installed via Snap and runs into issues when it tries to restart via systemctl. I would avoid installing via Snap and stick to the system packages.
Some stuff seems "left over" in the role, e.g. vars for
matomo
. It would be good to look at what vars are still used and what is essential. Perhaps also thesentry
stuff is unused?For a role with this level of complexity (several sub-roles, vars which change the execution flow), I would recommend setting up routine Molecule testing so the role can be guaranteed to be working.
You have several things to test already: Ansible version, Ruby version, Node version and Ubuntu version. And in between all the system dependency versions, e.g. Passenger. Without regular testing, this role can become unreliable.
The text was updated successfully, but these errors were encountered: