https://lists.fedoraproject.org/archives/list/kernel@lists.fedoraproject.org/thread/IE4Z4UWEAWD66US5MH7UGMV54P3BI2BN/

We used to have a hardware database called Smolt. It was largely unmaintained and had various problems. But it provided valuable data. The biggest problem was that changes were very hard. It was fairly easy to collect data. But all access to the data required the development of full web pages. There was no data API. So interest in the project waned.

In response to this, I created a project called Census. Census is a data collection and query API service. It makes it easy to collect data (plugin directory with executables that return JSON) and store it (server-side JSON parsing plugins). But most importantly, once the data is in Census it can be queried by anyone. You can submit a query and the result is returned to you in JSON. There is also a minimal web UI which serves as a development environment for crafting queries.

The goal is that domain experts can write plugins for the data they wish to collect and write queries to analyze that data. To that end, Census itself intends to be a minimal service that others can build upon.

Census currently requires MongoDB. But we'd like to change to PostgreSQL. Our requirements are as follows:

1. PostgreSQL master server
2. 0..n PostgreSQL read-only mirror query servers (to scale)
3. 1..n Web Front Ends (to scale); preferably OpenShift S2I

Load balancing would not be initially needed in the development environment. But as we scale it up, we may want to load balance across the web front ends and pgsql instances.

We do not require additional packages. Though we would require Python3 in the web front ends.