I'm trying to use failover option for Galera Cluster.
URL is like this:

jdbc:mysql:failover://192.168.33.11,192.168.33.12,192.168.33.13/test?autoReconnect=true

And when failover occurred in a transaction with connection to 192.168.33.13, the driver threw an exception with this message:

Communications link failure with primary host 192.168.33.11:3306. unexpected end of stream, read 0 bytes from 4.  Driver as successfully reconnect connection

But "192.168.33.11" is not the host which caused "communications link failure".
It should be "Communications link failure with primary host 192.168.33.13:3306".
"192.168.33.11" is the new, active master host.
(I have checked it with query log.)

I think the cause of this problem is in FailoverProxy#handleFailover().
The message above is created in listener.throwFailoverMessage(),
and in throwFailoverMessage(), the failed host address is retrieved from currentProtocol.getHostAddress().
However, currentProtocol is already updated in listener.handleFailover().

  FailoverProxy#handleFailOver()

  -> AbstractMastersListener#handleFailover()

  -> MastersFailoverListener#primaryFail()

  -> MastersFailoverListener#reconnectFailedConnection()

  -> MasterProtocol#loop()

  -> MastersFailoverListener#foundActiveMaster()

I found this problem with multi master (MastersFailoverListener),
but it might happen with master/slave replication (MastersSlavesListener)
because it has very similar implementation.