Put replication monitoring in a downtime for master2, slave1, slave2, slave3
"STOP SLAVE" on slave1, slave2, slave3
stop slave on master2. Check that Exec_Master_Log_Pos on master2 is later than on slave1, slave2, slave3. If not, start slave on slave1/2/3 and do step 2 again.
run the following on slave1, slave2, slave3, using the Master_Log_File and Exec_Master_Log_Pos from SHOW SLAVE STATUS on master2 (after the slave is stopped):
At this point, SHOW SLAVE STATUS run on master2, slave1, slave2, and slave3 should have the same Master_Log_file and Exec_Master_Log_Pos. If not, repeat steps 3-4.
Using the binlog file and position from SHOW MASTER STATUS on master2, run the following on slave1, slave2 and slave3:
After all the slaves are started and successfully slaving master2, run SLAVE START; on master2.
Wait for monitoring to clear up and remove the downtimes.