Migrating a network of replicating servers

The first time that the master server is started, it migrates the information in the directory that controls replication. The entries with objectclass replicaObject under cn=localhost are replaced with entries used by the new replication model (for more information, see Replication). The master server is configured to replicate all the suffixes in the directory. The agreement entries are created with the attribute ibm-replicationOnHold set to true. This allows updates made to the master to be accumulated for the replica until the replica is ready.

These entries are referred to as the replication topology. The new master can be used with replicas running prior versions; data related to the new functions will not be replicated to the back-level servers. It is necessary to export the replication topology entries from the master and add them to each replica after the replica server has been migrated. To export the entries, use the Qshell command line tool ldapsearch and save the output to a file. The search command is similar to the following:

ldapsearch -h master-server-host-name -p master-server-port \
 -D master-server-admin-DN -w master-server-admin-password \
 -b ibm-replicagroup=default,suffix-entry-DN \
 -L "(|(objectclass=ibm-replicaSubEntry)(objectclass=ibm-replicationAgreement))" \
     > replication.topology.ldif

This command creates an output LDIF file named replication.topology.ldif in the current working directory. The file contains only the new entries.

Note:
Do not include the following suffixes: Include only user-created suffixes.

Repeat the command for each suffix entry on the master, but replace ">" with ">>" to append the data to the output file for subsequent searches. After the file is complete, copy it to the replica servers.

Add the file to the replica servers after they have been successfully migrated; do not add the file to servers running previous versions of the directory server. You must start and stop the server before you add the file.

To start the server, use the Start option in iSeries Navigator. For more information, seeStart/stop the Directory Server.

To stop the server, use the Stop option in iSeries Navigator. For more information, see Start/stop the Directory Server.

When you add the file to a replica server, be sure that the replica server is not started. To add the data, use the Import File option in iSeries Navigator.

After the replication topology entries are loaded, start the replica server and resume replication. You can resume replication in one of the following ways:

To determine which replica agreement DN corresponds to a replica server, look in the replication.topology.ldif file. The master server will log a message that replication has started for that replica and a warning that the replica server's ID in the agreement does not match the replica's server ID. To update the replica agreement to use the correct server ID, use Replication Management in the Web administration tool, or the command line tool ldapmodify. For example:

ldapmodify -c -h master-server-host-name -p master-server-port \
 -D master-server-admin-DN -w master-server-admin-password
dn: replica-agreement-DN
changetype: modify
replace: ibm-replicaConsumerID
ibm-replicaConsumerID: replica-server-ID

You can enter these commands directly on the command line, or you can save the commands in an LDIF file and supply them to the command with the -i file option. Use End Previous Request to stop the command.

Migration for this replica is complete.

To continue to use a replica running a previous version, it is still necessary to resume replication using the command line tool ldapexop or Replication Management in the Web administration tool for that replica. If a replica running a previous version is migrated later, use the command line tool ldapdiff to synchronize the directory data. This will ensure that entries or attributes that were not replicated are updated on the replica.