Cluster Recovery

Introduction

What is Cluster Recovery? Cluster recovery requires understanding broker restarts, partition leadership, ISR, controller state, disk replacement, and client retry behavior.

Understanding the topic

What happens — Cluster Recovery:

• Identify topic, partition, offset scope.
• Check broker health and ISR.
• Inspect consumer errors and deploys.
• Remediate: scale, DLT, replay, or fix schema.

Visual explanation

Pipeline view:

Topic partitions
  ↓
replicas across brokers
  ↓
ISR health
  ↓
disk + network + controller
  ↓
SLO dashboards

Step-by-step explanation

1. Alert — Lag, URP, error rate.
2. Diagnose — Partition-level view.
3. Fix — DLT, scale, rollback.
4. Postmortem — Document root cause.

Informative example

Example:

kafka-topics --bootstrap-server localhost:9092 --create --topic cluster-recovery --partitions 6 --replication-factor 3
kafka-console-producer --bootstrap-server localhost:9092 --topic cluster-recovery
kafka-console-consumer --bootstrap-server localhost:9092 --topic cluster-recovery --from-beginning
kafka-consumer-groups --bootstrap-server localhost:9092 --describe --group cluster-recovery-service

Execution workflow

Best practices

• Use replication factor 3 for critical topics.
• Set min.insync.replicas with acks=all.
• Alert on under-replicated partitions, offline partitions, disk usage, and lag.
• Document topic ownership and cleanup policy.

Common mistakes

• Changing partitions without understanding ordering impact.
• Ignoring hot partitions caused by skewed keys.
• Letting retention grow until disks become the outage.

Summary

Cluster Recovery — Operate Kafka through SLOs: durability, availability, latency, lag, disk headroom, and recovery time.