> ## Documentation Index
> Fetch the complete documentation index at: https://docs.risingwave.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Serving CPU load pressure

> Diagnose and mitigate Serving Node CPU load pressure in RisingWave Cloud.

This alert indicates that Serving Node CPU utilization has remained above the configured threshold. Sustained CPU pressure can increase query latency or reduce query throughput.

## Assess serving health

1. Open the cluster's [Metrics page](/cloud/check-status-and-metrics) in the RisingWave Cloud portal. Select a time range that starts before the alert was triggered and includes the current time.

2. Compare the following metrics with the cluster's normal baseline and with your application-level query metrics.

   | Metrics tab  | Metric                                                                    | Look for                                                                                                                                                                      |
   | ------------ | ------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
   | **Resource** | **CPU Utilization**                                                       | Node CPU utilization that remains elevated during the alert period. Use the query metrics and application signals below to determine whether serving performance is affected. |
   | **Query**    | **Query Local Average Latency** and **Query Distributed Average Latency** | An increase in query latency that coincides with the CPU pressure.                                                                                                            |
   | **Query**    | **Queries Per Second**                                                    | A sustained increase in query volume, or an unexpected decrease accompanied by application-side latency or failures.                                                          |

3. Check application-level signals for increased latency, timeouts, or failed queries.

## Scale serving resources

If serving performance is affected or the cluster does not have sufficient CPU headroom, [scale the cluster](/cloud/scale-a-project-manually):

1. Open the cluster from the [RisingWave Cloud dashboard](https://cloud.risingwave.com/dashboard).
2. Click **Scale**.
3. Open the custom configuration for **Serving**.
4. Apply one or both of the following changes:
   * To scale up, increase the RWUs allocated to each Serving Node.
   * To scale out, increase the number of Serving Nodes.
5. Review and apply the new configuration.

## Confirm recovery

Monitor **CPU Utilization** and the query metrics until they return to their normal baselines. Confirm that application query latency and error rates have recovered and that the alert has cleared.

If CPU pressure or query degradation continues after scaling, contact the [RisingWave Cloud support team](mailto:cloud-support@risingwave-labs.com). Include the cluster name, alert trigger time, metric screenshots, and the serving-resource changes already applied.

## Related topics

* [Check status and metrics of clusters](/cloud/check-status-and-metrics)
* [Scale a cluster manually](/cloud/scale-a-project-manually)
* [Monitor messages and alerts](/cloud/monitor-messages-and-alerts)