Design, route, and scale database read replicas with confidence. Over eighty in-depth guides covering topology design, connection pooling, load balancing, ORM routing, replication lag, caching, failover and promotion, routing correctness, and read-tier observability.
Built for backend engineers, DBAs, SREs, and platform architects who need production-grade patterns — not toy examples. Every guide includes trade-off matrices, annotated configuration, failure-mode analysis, and the verification step that proves the change worked.
Four Core Domains
From topology design and query-level routing to lag management, caching, failover, and fleet observability — twenty-four in-depth topic guides and over fifty step-by-step runbooks, grounded in real production systems.
Master topology design, synchronous vs asynchronous replication, consistency models, multi-region deployment, and the promotion mechanics that decide what a failover costs you.
Implement read/write splitting at the proxy layer, configure pgBouncer, build ORM middleware routes, balance reads across the pool, and prove the split is correct before it ships.
Detect and handle replication lag in real time, implement fallback strategies, route queries by freshness, and keep caches and materialized views from compounding staleness.
Instrument replica health with Prometheus, build Grafana dashboards for replica fleets, alert on lag and pool saturation, analyze query performance, and turn routing decisions into traces and SLOs.
Start Here
New to read replicas? Work through these seven guides in order for a solid production foundation.
Latest Additions
Five new in-depth guides covering the areas teams reach once the basic split is working: promotion mechanics, replica-pool balancing, routing correctness, layered staleness, and read-tier SLOs.
The one operation where a read replica stops being a read replica: fencing, candidate selection, timeline divergence, and repointing the routing tier.
Explore section →Adding replicas only helps if reads actually spread across them. Selection algorithms, lag-derived weights, zone preference, and clean draining.
Explore section →Safety, completeness and coherence — with a CI fixture, a standing digest audit, and a canary whose criteria include a replica-share floor.
Explore section →Staleness composes. One budget per read path, invalidation that survives the lag window, and materialized views that do not cancel replica queries.
Explore section →From node health to read quality: span attributes that explain a routing decision, a staleness SLO beside the latency one, and a weekly report that drives decisions.
Explore section →Every guide includes decision matrices comparing approaches across latency, complexity, and operational overhead.
Copy-ready YAML, SQL, Python, and INI configurations drawn from real deployment patterns.
Explicit failure patterns, root causes, and mitigations for every architectural decision.
Patterns covering PostgreSQL, MySQL, ProxySQL, PgBouncer, HAProxy, SQLAlchemy, Django, Prisma, Prometheus, and Grafana.