Performance Degradation After Adding a Second Compute Node in RisingWave (Slow Streaming & MV Creation Timeout)

[MKRyuto]

Hi everyone,

I’m experiencing an issue after adding a second compute node in RisingWave.

Initially, I was running RisingWave with one compute node, and everything worked smoothly. However, after adding a second compute node, I started seeing inconsistencies and performance degradation.

Issues observed:

1. Streaming performance becomes noticeably slower, especially when frequent data updates are occurring.
2. Creating a materialized view with large datasets becomes extremely slow and often ends up timing out.
3. Overall query performance feels worse compared to when I only had a single compute node.
4. It seems like the compute nodes sometimes fall out of sync or take longer to align state during processing.

My setup:

• Using docker-compose-distributed.yml
• Two compute nodes (compute-node-1 and compute-node-2)
• Each compute node has sufficient resources (32 GB RAM per node)
• Memory parameters were configured according to the documentation
• With only one compute node, all streaming and MV operations run without issues

Questions:

• Is there any special configuration required when running more than one compute node?
• Do I need to adjust anything related to state store, barrier alignment, or shuffle configurations for multi-compute setups?
• Has anyone experienced slow MV creation or timeouts after scaling out compute nodes?

Any help, insights, or recommendations would be greatly appreciated. 🙏

Thanks!