After discussing internally on the #reading-infrastructure-team, primarily with @Tgr , we have the following likely possibilities:
1. Use a new DB cluster on the standard DB servers.
2. Use one of the existing non-mediawiki DB clusters (m1-m5, which are used for all kinds of stuff from Phabricator to the scholarship review app)
3. use a cross-wiki MediaWiki DB cluster (x1, which is currently mostly used by Flow, or just stick a DB into one of the clusters which are normally used for per-wiki tables - e.g. the CentralAuth database is hosted on s7)
We also have 2 other possibilities but think they are less desirable/likely:
4. use a separate database server
5. use wiki-specific tables (s1-s7, one table per wiki)
So far, hosting this on s7 (option 3) seems like the best option. Why?
- It gives us the options to link devices to user accounts if we ever decide to do that.
- It is non-wiki specific (and devices and browsers are also non-wiki specific)
- We don't have to setup a new DB (not terrible, but more work)
So my questions:
- Did we miss any options?
- Are there any downsides/upsides we missed to any of the options?
- Does this logic seem reasonable?
I haven't done so yet, but will be updating the technical plan with this info shortly:
https://docs.google.com/document/d/1nVW3XC7PO3XDKRz72fCFDoZ-CA_lrhzNekUldPn5wFA/edit?pli=1#heading=h.xanehm3k39vh