Handling Simultaneous Address Changes
If an address change occurs simultaneously for the master replica and a slave replica, the master and slave will not be able to reach each other while both are trying to notify the other of the changed address. To avoid this problem, make sure the address change of one replica (either master or slave) is propagated to all replicas before the other address is changed. Make one address change. Then, use the dcecp registry show -replica command to view the replica list at both the master site and the slave replica site. When the new address is displayed, on both replica lists, it is safe to proceed with the next network address change.
If you are unable to prevent simultaneous network address changes for the master and a slave, the only way to restore communication between the master and slave is to delete the slave, then recreate it. Delete the slave by using one of the following methods, depending on your circumstances:
· If you anticipate a simultaneous address change, while the master and slave are still communicating, use the set command to bind to the master and then the dcecp registry delete command to delete the slave replica.
· If secd is running at the master and slave sites, but the master and slave are not communicating, first use the set command to bind to the slave and then the registry destroy command to destroy the slave. Then use set to bind to the master and the registry delete -force command to remove the replica list entry for the slave.
· If secd is not running at the slave site or if you are unable to bind to the slave site, use the procedure for recreating a replica described in Troubleshooting Procedures.