Replication Operations Report
ReplMon · transactional replication health, incidents and repairs
1. Executive Summary
2. Latency Against the SLO
Tracer tokens posted every minute under continuous OLTP load, per subscriber. Dashed line: the latency SLO.
3. Where the Time Goes
Median per hop. Publisher → distributor is the Log Reader; distributor → subscriber is the Distribution Agent and the network.
4. Incidents
Show incidents
| # | Kind | Subscriber | Detected | Diagnosis | Action | Resolved |
|---|
5. Rows Repaired
Counted by type, table and subscriber. Every repaired row is in the CSV beside this report; the most recent are shown in full with the values before and after.
| Type | Table | Subscriber | Rows | First (UTC) | Last (UTC) |
|---|
Show repaired rows
6. DBA Guardrail
Every repair above was row-level and driven by the publisher. Before closing the incident:
- Find who changed subscriber data outside replication. Subscribers here are read-only; a missing, stray or changed row has a cause.
- Never clear these errors with
-SkipErrors: the agent goes green and the subscriber stays wrong. - Broad damage (many rows, an expired subscription) calls for reinitializing the affected articles, not repeated repairs.