I totally agree with you but this is not immediate. We use repmgr to handle the failures and the DBAs have setup 6 retries with 10 seconds delay between each retry.
So during 1 minute I'm not able to serve read queries even if my slave is available, the preferMaster option would allow me to do it very easily.
I have a setup with 1 master and 1 slave (hot standy to serve read queries if necessary). I would like the JDBC driver to talk preferrably to the master (to be sure to get the freshest data). If the master is not available I would like to send the queries to the slave.
The documentation mentions the following values for the parameter targetServerType: any, master, slave, preferSlave. I'm surprised there is no preferMaster option which would allow me to implement easily my requirement.
normally, if the master is not available, that means its failed, and your HA cluster management will promote the slave to be the new master, fencing off the old master until it can be manually reset and restored as the new slave.