#reliability

5 posts.

Why Kafka Transactions Cannot Commit with Your Database Yet: KIP-939 and 2PC Participation

Learn what Kafka transactions can commit today, why databases and external systems remain outside that boundary, and how KIP-939 proposes to let Kafka participate in externally coordinated 2PC.

kafkatransactionsexactly-oncearchitecturereliability

Kafka EOS: What Is Actually Exactly Once?

A precise look at what Kafka Exactly-Once Semantics protects across producer retries, transactions, and consumer offsets—and why keys, headers, and event IDs are not broker-side deduplication rules.

kafkaexactly-oncetransactionsconsumersreliability

Active-Active DR: Recovery, Reconciliation, and Failback

Learn how to prove a recovery checkpoint, reconcile uncertain data, and execute failback safely after an Active-Active DR failover.

architectureactive-activedisaster-recoveryreliabilityoperationskafka

Home Region Writes: Keeping Database and Kafka Ownership Aligned

Learn how to enforce Home Region ownership across request routing, database transactions, Outbox, Kafka, epoch fencing, and ownership transfer in a multi-region service.

architectureactive-activehome-regiondisaster-recoverykafkareliability

Why Kafka Fits Between Outbox and Inbox Patterns

This post explains why Outbox and Inbox patterns matter in asynchronous MSA flows, and why Kafka is a strong fit as the message path between them.

kafkamsaarchitecturemessagingreliability