Using LOAD, UNLOAD, and BEGIN WORK WITHOUT REPLICATION

When you add a new participant to an existing replicate, you can unload and load data without replication.

About this task

Procedure

To add a new participant to an existing replicate
  1. Add the server delta to the Enterprise Replication domain.
    For example: 
    cdr def ser -c delta -I -S g_alpha g_delta

    At the end of this step, all servers in the replication environment include information in the syscdr database about delta, and delta has information about all other servers.

  2. Add delta as a participant to replicate zebra.
    For example: 
    cdr cha rep -a zebra "P dbname@g_delta:owner.table1" \
"select * from table1"

    This step updates the replication catalog. The servers alpha, beta, and gamma do not queue any qualifying replication data for delta because the replicate on delta, although defined, has not been started.

  3. Suspend server delta on alpha, beta, and gamma. 
    cdr sus ser g_delta g_alpha g_beta g_gamma

    As a result of this step, replication data is queued for delta, but no data is delivered.

  4. Start replication for replicate zebra on delta. 
    cdr sta rep zebra g_delta

    This step causes servers alpha, beta, and gamma to start queuing data for delta. No data is delivered to delta because delta is suspended. Then, delta queues and delivers qualifying data (if any) to the other servers.

    Do not run any transactions on delta that affect table table1 until you finish the synchronization process.

  5. Unload data from table table1 using the UNLOAD statement.
  6. Copy the unloaded data to delta.
  7. Start transactions with BEGIN WORK WITHOUT REPLICATION, load the data using the LOAD statement, and commit the transactions.
  8. Resume server delta on alpha, beta, and gamma.
    cdr res ser g_delta g_alpha g_beta g_gamma
    This step starts the flow of data from alpha, beta, and gamma to delta.

    At this point, you might see some transactions aborted because of conflict. Transactions can abort because alpha, beta, and gamma started queuing data from delta in step 4. However, those same transactions might have been moved in steps 5 and 7.

Results

You must declare replication on server delta and then immediately suspend replication because, while you are preparing the replicates and unloading and loading files, the other servers in the replicate (alpha, beta, and gamma) might be collecting information that needs to be replicated. After you finish loading the initial data to delta and resume replication, the information that was generated during the loading process can be replicated.