It appears that your document is an in-depth technical assessment of StarRocks' database system, focusing on its reliability and performance issues. The analysis highlights several critical problems:
-
Memory Management Issues:
- BE nodes frequently crash due to out-of-memory (OOM) errors.
- Memory tracking is unreliable; reported memory usage can be vastly inaccurate.
- Compaction tasks can enter infinite retry loops, causing system instability.
-
Query Result Integrity Problems:
- Iceberg cache serves stale data without any error or warning.
- Arrow Flight silently drops rows from results under certain conditions.
- Non-deterministic functions like
RAND()are incorrectly pushed down, leading to empty result sets.
-
Real-Time Ingestion Reliability Issues:
- StreamLoad operations can freeze the entire cluster.
- Flink connector fails with HTTP 404 errors upon pipeline restarts in shared-data mode.
- Kafka Connect jobs get stuck in transaction states for extended periods, causing delays.
- INSERT performance degrades significantly as tables grow.
-
Query Performance Instability:
- Cost-Based Optimizer (CBO) selects inefficient plans when statistics are stale or inaccurate.
Read the full article at DEV Community
Want to create content about this topic? Use Nemati AI tools to generate articles, social posts, and more.

![[AINews] The Unreasonable Effectiveness of Closing the Loop](/_next/image?url=https%3A%2F%2Fmedia.nemati.ai%2Fmedia%2Fblog%2Fimages%2Farticles%2F600e22851bc7453b.webp&w=3840&q=75)



