[PATCH] db: add the ability to have a seperate write database

[Axel Gembe]

This patchset makes Trac able to have a seperate database for writing which is useful in master->slave replication scenarios.

[Axel Gembe]

[PATCH 1/3] db: add the ability to have a seperate write database

[Axel Gembe]

[PATCH 1/3] db: add the ability to have a seperate write database

[Axel Gembe]

[PATCH 2/3] db: distinguish between read and write access

[Axel Gembe]

[PATCH 3/3] db: make trac-admin use database_write

[nkantrowitz]

To summarize my comments from the email thread: Trac stores things outside the database anyway, so database replication really isn't useful given the added complexity. -1.

[Andrew Shearer <ashearerw@…>]

This would still help with database load balancing, and would also help if the primary database server goes down. As Noah said, it doesn't help the situation much if the web server itself goes down, but that's outside the scope of this ticket. With some installations already having database master/slave replication set up for other applications, and a database server sometimes being a bottleneck, this looks like a net win. +1.

[nkantrowitz]

If the concern is just the database, Postgres already offers a number of HA solutions (which would include replication) and I assume MySQL still has its clustering mode.

[rblank]

The patch is huge, breaks some of the most basic interfaces used in many plugins as well, and according to comment:3, the same effect can be achieved outside of Trac.