General troubleshooting approach

When you encounter performance problems in RisingWave, follow these steps to systematically identify the root cause:

1. Check overall throughput and latency: Start by examining the key performance indicators described in Monitoring and metrics. Are your queries exhibiting high latency or low throughput? What are the specific values you’re seeing, and how do they compare to your expected performance? High latency can often indicate issues with query structure or resource constraints.

2. Identify backpressure: Backpressure, where a downstream component slows down an upstream component, is a common indicator of bottlenecks. Workload analysis provides a detailed explanation of backpressure and how it works in RisingWave. For specific steps on identifying backpressure, see Troubleshooting high latency.

3. Examine resource utilization: If backpressure is present, or if overall performance is poor, investigate resource utilization. Monitoring and metrics provides an overview of key resource metrics, including CPU usage, memory usage, and cache performance. For a detailed guide to identifying specific resource limitations (e.g., CPU bottlenecks, state access bottlenecks, etc.), see Specific bottlenecks.

4. Check for errors: Examine RisingWave’s logs for any error messages that might indicate the source of the problem.

5. Simplify the query: Try to remove parts of the SQL query to isolate which part is causing the issues.

By following these steps, and using the detailed guidance in the linked sections, you can effectively diagnose and resolve most performance problems in RisingWave.