HP Integrity NonStop H-Series - Manual (Page 51)

SDR Operations
HP NonStop SQL DDL Replicator User’s Guide —545799-006
4-11
SDR and RDF Takeover
When executing replicated DDL, SDR reports SQL errors in the EMS log and waits for
the operator to take action. See
Controlling Replication in SDR Monitoring and Control .
If the DDL statement alters or drops a table that does not exist on the backup, the
operation is discarded and the RDF updaters are restarted.
When executing the DDL for a CREATE operation, SDR assumes the user identity that
executed the primary system DDL. Thus, the same user will own both primary and
backup SQL objects. You can configure the user ids that SDR should use for CREATE
operations. See
Setting CREATE Ownership and Security on page 3-5 for details.
SDR executes the DDL for other types of access as the SUPER user, just as RDF
does for data replication. You can configure SDR to perform replicated DDL operations
other that CREATE as the user who performed the operation on the primary. See
Setting the User ID for Replicated DDL Operations on page 3-6 for details.
Restarting RDF Update
Once the backup DDL is successfully executed, or canceled by the operator, SDR
sends RDF a message to restart its updaters. SDR then waits for the next RDF
“Update NSA Stopped” state.
SDR and RDF Takeover
SDR operation in a planned or unplanned RDF takeover is completely automatic; no
operator intervention is required.
An RDF takeover can occur only if the primary system has failed (or is unavailable on
the Expand network) or if RDF has been stopped on the primary system. Either of
these events will cause the SDR updaters started from the primary system to shut
down.
When you issue the RDF TAKEOVER command, the SDR monitor on the backup
system immediately starts an SDR updater for that RDF configuration. If there are any
SDR-captured DDL operations in the remaining RDF image trail, RDF will shut down
on each SRU audit record, SDR will execute the DDL operation and then restart the
RDF takeover. Eventually, RDF will enter the “Takeover Complete” state and both RDF
and SDR will shut down.
Note that, for complex RDF topologies, the SDR monitor can start SDR updaters for
both primary and takeover RDF configurations on the same system.
Distributed Table Replication
If a table or index has partitions and indexes on remote systems, the DDL capture
process has written a SRU audit record to the audit trail of each system that has a
partition or an index of the table. On each one of those systems, RDF is expected to
enter the “Update NSA Stopped” state for the same file name. (Actually, each SRU file