Disaster Recovery Options running Apache Kafka in Kubernetes
summary
Concrete choices and tests rather than marketing: deploy stretch clusters across three regions with nonoverlapping CIDRs and at least two brokers per region for near‑zero RTO, use replicator / mirror maker for async near‑zero RPO or cluster linking to preserve offsets, run 3–5 ZooKeeper nodes (or migrate to KRaft), keep ISR >=2–3 and two observers per region, bind SSD volumes with Retain reclaim policy, configure kube‑DNS/loadbalancer routing and node/pod affinity, automate with a Kubernetes operator and Confluent tooling — observer promotion and failover tests show no data loss.
AI-written summary · about
chapters (14)
- 0:00Start
- 0:25Introduction about Myself
- 4:06What Kafka Is
- 5:07Resiliency in Kafka
- 6:42Partitioning Model in Kafka
- 11:33Replica Placement Policy
- 18:15The Differences between Deploying Kafka on Kubernetes versus on-Prem Disaster Recovery
- 18:45Namespaces
- 19:16Kubernetes Operator
- 22:04Resolving Networking between Kafka Clusters on Kubernetes
- 24:05Define Ordinality in Kafka
- 28:07Zookeeper Configuration
- 34:15Health Checks
- 34:54Storage Class Types
related talks
24mMake Your Kafka Cluster Production-Ready
47mRunning Kafka in Kubernetes: a practical guide
46mKubernetes Stateful Application Disaster Recovery with Zero Data Loss
13mScaling Kafka on Kubernetes
36mDisaster Recovery in Action with Kafka and Strimzi
36mKafka Without Disks: Making Streaming Brokers Disposable on K8S
29mKafka on Kubernetes: Does it really have to be the hard way?
27mRunning Kafka in Kubernetes a practical guide