Exploratory Sprint: Distribution of Meltano

Longer term, we need to put thought into how we want to approach distribution of Meltano. This issue captures the things we need to think about.

Current Sit Rep

Right now, we did something very simple with DigitalOcean one-click so they can see the UI and test it out. But in order to make this performant for a real production user we will need to identify the requirements.

We did this to gain MAUI by reducing friction, but in order to retain MAUI we will have additional work to do. Within a few months, we will probably have some users who will have challenges with performance, upgrades, etc.

Open Questions

• Meltano is a Python application, what components do you need to run? What ships with the Meltano package versus the image?
• How much do we make you do yourself, versus how much do we do for you?
• How do we avoid taking on too much maintenance?
• What are the principles upon which we decide or "pick our fights" in terms of what problems we will solve in the deployment sphere (e.g. we would probably solve syncing production vs. dev environment but we probably would not solve setting up a reverse proxy -- we can have instructions/images but not a silver bullet)?

Known Issues

• Setting up HTTPS is difficult with one-click
• Users struggle to set up Postgres for themselves
• Setting up Postgres for users locally is problematic, but doing this for a cloud native scenario could be better

Next Steps

• Continue to assess how much friction there is in the installation process
• Document issues users report related to this step in the process
• Figure out where we will draw the line on what we will solve for users, versus what we will document in terms of instructions for them to do themselves