RDF/IMP and IMPX System Management Manual (RDF 1.3+)

Online Database Synchronization
Compaq NonStop™ RDF/IMP and IMPX System Management Manual522204-001
7-18
Phases of Online Database Synchronization
Phases of Online Database Synchronization
Online database synchronization, whether for entire databases or selected volumes,
occurs in two phases for both the extractor and all affected updaters.
Extractor Phases
The extractor phases of online database synchronization are as follows:
Phase 1, Part 1
The extractor has received the STOP SYNCH message (indicating the load or backup
operation is complete) and has reached the first TMP control point record in the audit
trail that was generated after the load or backup operation completed. At this time, the
extractor begins building a list of all transactions that might have been started during the
create/load or backup operation. Upon completion of phase 1, part 1, the extractor logs
message 766.
Phase 1, Part 2
The extractor has reached the next TMP control point record in the audit trail and now
has a list of all transactions that might have been started during the create/load or backup
operation. The extractor generates a synch-phase-1-complete record, and then continues
its normal operation.
Upon completion of phase 1, part 2, the extractor logs message 767.
Phase 1, Part 3
When the extractor encounters the next TMP control point record and the list is empty,
the extractor generates a special synch-complete image record (which is stored in all
image trails by the receiver). All transactions that might have been started during the
create/load or backup operations are now finished.
Upon completion of phase 1, part 3, the extractor logs message 768.
Phase 2
Phase 2 completes when the extractor is certain the synch-complete image record has
been successfully written in all image trails, and the extractor’s restart location is at a
point in the audit trail following the TMP control point record associated with the
completion of phase 1, part 3, above.
Upon completion of phase 2, the extractor logs message 782.
Note. TMP control points are generated as the result of transaction activity. For high rates of
transaction activity, the TMP control points might be only 1 or 2 minutes apart. For lower rates,
they might be 5 to 10 minutes apart. On a completely idle system, TMP control points can be
approximately 30-60 minutes apart. Therefore, if your applications and TMF are idle or nearly
so, it could take an hour or more for the extractor to encounter two TMP control points.