Does Nuxeo support routing read-only database transactions to PostgreSQL 9 HotStandby servers as a way to improve the overall scalability of a Nuxeo deployment?
That's something I would really like to be able to do, but it's not possible at the moment.
You would need to know at the load-balancer level if the HTTP request is going to correspond to a read-only transaction or not. And we don't have any global convention for that at the moment, as some read-only transactions are using POST methods (due to JSF), and some GET methods are used to trigger possibly anything in Nuxeo, which could potentially write to the database.
If you code a JAX-RS-based application and strictly adhere to a convention that allows you to decide from the HTTP request if there will be database writes or not, then you could go a long way toward what you're suggesting.
Is it possible to identify some high-value operations and configure them to be handled using a "read-only" data source? Some WebDAV and CMIS read-only operations might fit this category. Also, the Web UI Advanced Search seems like another high-value candidate.
Yes probably a subset of operations could be identified and bound to a specific datasource.
