10.8.3 – Fault Tolerance

Module: Section 10 – Kafka Ecosystem & Tools
Section: 10.8 – ksqlDB in Production
Goal: Understand ksqlDB fault tolerance mechanisms

State Replication

Automatic via changelog topics:

Server 1 (active): Processes partition 0 → Writes to state store
                                         → Writes to changelog topic
Server 2 (standby): Reads changelog topic → Replicates state store

If Server 1 crashes: Server 2 takes over partition 0 (state already replicated)

Recovery Time

Factors:

• State store size
• Changelog topic lag
• Disk I/O speed

Typical: 30-120 seconds for 1 GB state store

Rebalancing

Triggers:

• Server added/removed
• Server crash
• Network partition

Impact: Temporary query pause during rebalancing

Key Takeaways

1. Changelog Topics: Automatic state replication
2. Standby Replicas: Ready to take over
3. Recovery: 30-120s typical
4. Rebalancing: Temporary pause

What's Next?

Next Module: 10.8.4 – Operational Best Practices