Troubleshooting Guide

Common operational failure modes and remediation runbooks for Beluga.

1. High Kafka Consumer Lag

• Cause: Flink operator bottleneck or partition skew
• Remediation: Scale Flink task parallelism and increase topic partition counts

2. Flink Checkpoint Timeouts

• Cause: MinIO storage I/O contention or large RocksDB state payloads
• Remediation: Verify incremental checkpointing is enabled; audit MinIO disk latency

3. Trino Query Memory Exhaustion

• Remediation: Tune query.max-memory-per-node and enforce partition pruning in SQL queries