replication-monitor

name	replication-monitor
description	Monitors database replication health (streaming, logical, multi-region) and diagnoses lag, conflicts, and failover readiness.
allowed-tools	Read, Write, Bash
effort	low

Replication Monitor

When to activate

When replication lag exceeds acceptable thresholds, during failover testing, when setting up new replicas, or during routine weekly health checks.

When NOT to use

Skip if replication lag is under 1 second and all replicas report healthy. Skip for single-node development databases.

Instructions

1. Check replication status: pg_stat_replication for PostgreSQL, SHOW REPLICA STATUS for MySQL.
2. Measure lag: write lag (sent vs flush), replay lag (flush vs replay), byte lag.
3. Identify lag sources: large transactions, DDL operations, network saturation, disk I/O on replica.
4. Test failover: simulate primary failure, measure time-to-promote, verify data consistency.
5. Check for replication conflicts: long-running reads blocking WAL apply, hot standby feedback issues.

Example

REPLICATION STATUS: prod-primary → 2 replicas

Replica: prod-replica-1 (us-east-1)
  State: streaming ✅
  Write lag: 120ms | Replay lag: 180ms
  Byte lag: 2.4MB
  Conflicts: 0

Replica: prod-replica-2 (us-west-2)
  State: streaming ✅
  Write lag: 450ms | Replay lag: 620ms    ⚠️
  Byte lag: 18MB
  Conflicts: 3 (long-running analytics queries)

Action: Cancel 3 conflicting queries on replica-2 or increase
        max_standby_streaming_delay to 30s.