The upshot is that we want to minimize the number of ad hoc deployment strategies we use to maintain legacy applications, e.g., manually spinning up containers, installing dependencies outside of our version controlled server config, as these only further technical debt.
In an ideal world, production apps would either be deployed to their own servers using our legacy zero-downtime strategy, or (preferably) deployed using our latest Heroku strategy. The in-between areas are danger zones! If we deviate from these practices, it should be clearly and prominently documented.
Documentation request
Ported from https://github.com/datamade/devops/issues/124.
The upshot is that we want to minimize the number of ad hoc deployment strategies we use to maintain legacy applications, e.g., manually spinning up containers, installing dependencies outside of our version controlled server config, as these only further technical debt.
In an ideal world, production apps would either be deployed to their own servers using our legacy zero-downtime strategy, or (preferably) deployed using our latest Heroku strategy. The in-between areas are danger zones! If we deviate from these practices, it should be clearly and prominently documented.