Episode 3. GoldenGate. Day 3 of your 10 day sprint.

Welcome. DataGuard copies redo. GoldenGate copies committed change as logical replication, even across different engines. That difference is the whole episode.

Concepts. Extract captures from redo or logs, pump ships trail files, replicat applies to the target. Supplemental logging must be on and archive logs retained or capture misses rows. Lag is measured per process with LAG and INFO ALL. Conflicts come in three flavors: insert duplicates, update collisions, delete of a missing row.

Say this sentence: Every bi-directional table gets a conflict rule before go-live, or replication loops.

Scenario. Replicat lag grows on the reporting target after a batch load. Understand: which process lags, which tables, since when. Prepare: trail file growth, checkpoint age, target apply rate. Identify: a missing index on the target slowing apply versus source burst volume; check the apply plan first. Support: add the index or parallelize replicat, drain the backlog, add per-process lag monitors.

Say this bridge once. Say: GoldenGate CDC and DMS CDC share the same prerequisites, keys and source logging, so my conflict and lag discipline transfers directly to your RDS migration work.

Your Day 3 lab: design conflict resolution for one table, latest-timestamp versus trusted-source, and write the lag-monitor thresholds.
Teaser: Tomorrow, Episode 4, performance tuning. The AWR drill order you recite in your sleep.

Second scenario. A bi-directional pair starts ping-ponging updates. Understand: loop signature in trail growth. Prepare: conflict rules per table plus loop-detection tags. Identify: no rule on the hot table; add trusted-source immediately. Support: stop the loop, reconcile rows, publish the rule checklist.

Pause and say the four steps once. Understand. Prepare. Identify. Support. Say it again.

Close with your intro spine from memory. Ninety seconds, no notes. Done.