Semisynchronous replication

[Juan Comas]

Hi,
I'm trying to get the following replication process:

I have 2 computers.
Computer 1:
Master 1
Slave 1

Computer 2:
Master 2
Slave 2.

Slave 2 is a slave for Master 1 and
Slave 1 is a slave for Master 2.

Using heartbeat a virtual IP is assigned to the Computer 1 and when it fails, the Computer 2 starts accepting requests automatically. When the Computer 1 comes back.. heartbeat doesn't failback to Computer 1 until Computer 2 crushes..

As Slave 1 is hearing for Master 2 events, Slave 1 is getting synchronized. So if Computer 2 crushes.. Computer 1 should be ready to serve..
As I see things:
If wait enough for the Slave to be almost synchronized with the Master, there should be a moment when the Slave should have all events from Master in it's bin or in it's relay log. So a crush of Master 1 should be no problem cause the data is in the slave. The only thing that we should do is wait for the event to be processed from the relay in order to continue accepting requests..

Questions..
1) Is this topology/solution viable using semisynchronous replication?

2) When Computer 2 start receiving requests (because of a failover), Mysql waits for the relay log to be processed before attending any new requests so that the new request doesn't read old information?.



Thanks
J.J