NonStop SQL DDL Replicator User's Guide (Update 5)

SDR Operations
HP NonStop SQL DDL Replicator User’s Guide—545799-005
4-8
Replication Operations
mode, it is assumed that the primary system is unavailable; the SDR monitor on the
backup system manages the SDR updater.
EMS Messages
In contrast to RDF, each SDR component, including the SDR updater, logs all EMS
events to the system that it was started from. When RDF is in normal mode, all SDR
EMS events are sent to the primary system EMS collector. When RDF is in takeover
mode, all SDR events are sent to the backup system EMS collector. This maintains a
better time line of all SDR events and avoids problems comparing primary and backup
EMS logs.
Replication Operations
Replicating and executing DDL statements entails the following operations:
Creating the SDR Depot Files on the Backup
Processing Stop RDF Update Audit Records
Translating DDL
Executing DDL
Restarting RDF Update
Creating the SDR Depot Files on the Backup
The SDR depot files $vol.ZASDRnnn.SDRDEPOT contain the DDL statements
captured on the primary system. The creation of the depot files is not replicated to the
backup by RDF. When RDF first attempts to insert updates into the depot file before it
exists on the backup, it issues event 736. The SDR updater detects the event signaling
that RDF is looking for the file and creates the depot file.
Processing Stop RDF Update Audit Records
When the RDF Extractor processes an SRU audit record, it filters it based on the
associated file name. If the volume or subvolume is not protected by the RDF
configuration, the SRU record is discarded.
If the SRU record is accepted, it is sent to the RDF receiver who places it in every
updater image trail. If the SRU record is processed, it causes all RDF updaters to shut
down, with the backup database consistent up to the point of the associated DDL
statement. When all updaters have shut down, RDF issues an EMS 908 event. When
the SRU is associated with SDR, the filename will have the form
$vol.ZASDRnnn.AAmmmmmm where nnn is the primary Expand node number, and
mmmmmm is a unique sequence number assigned by SDR.
Note that the 908 shutdowns do not stop normal RDF extractor-to-receiver processing.
All audit produced on the primary database is still being transmitted and stored on the