Docs and scripts for Fedora Datanommer and the fools who love her
- Python 96.8%
- Shell 3.2%
| docs | ||
| iac | ||
| .gitignore | ||
| fnox.toml | ||
| README.md | ||
How to have an analytics datanommer
Process
- Have a database:
- Restore a full upstream backup (~17 hours)
- Congrats! You can't do very much because the messages are json in a text field 🙃. (This is
because some of the incoming json is >256M or has
\u0000JSON nulls in it.)
- Have a database you can actually query
- For an overview of all changes made, see: All changes to the public schema.
- To implement this on your own replica, see:
- Tuning (this can dramatically improve the performance of the rest of these steps!)
- Upgrade a PG13 datanommer to PG18 (~12 hours)
- Convert the json text to jsonb (~200 hours, 36x faster queries!)
- Convert the
topiccolumn to ltree representation (~20 hours, 2x faster queries)
- Have fun!