[Verse 1] When disaster strikes your clusters, cities burning down Active-passive holds the crown, standby waits around Primary region serves requests while backup sleeps in wait Mirror-maker replicates, no conflicts to debate [Chorus] Three patterns for survival when the networks fall apart Active-passive, active-active, choose your beating heart RPO measures data lost, RTO time to heal Monitor that replication lag, keep your clusters real [Verse 2] Active-active spreads the load, both regions running hot Conflict resolution needed when the same keys hit the spot Last-writer-wins is simple but data might get dropped Custom merge strategies help when logic can't be stopped [Chorus] Three patterns for survival when the networks fall apart Active-passive, active-active, choose your beating heart RPO measures data lost, RTO time to heal Monitor that replication lag, keep your clusters real [Bridge] Recovery point objective counts the seconds data's gone Recovery time objective tracks how long before we're on Lag metrics tell the story of how far behind we trail Prometheus alerts firing when replication starts to fail [Verse 3] Stretch clusters span geography but split-brain makes you weep Rack awareness spreads replicas, availability runs deep Cross-region network latency will slow your throughput down Bandwidth costs and vendor locks can make your budget frown [Chorus] Three patterns for survival when the networks fall apart Active-passive, active-active, choose your beating heart RPO measures data lost, RTO time to heal Monitor that replication lag, keep your clusters real [Outro] When earthquakes shake your data centers, floods wash servers clean These patterns keep Kafka flowing through disaster's darkest scene
← 2 Strimzi MirrorMaker 2 Configuration | 4 Lab: Multi-Cluster Replication →