> ## 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.

# Cluster health check failed

> Assess the impact of a failed health check and protect connected systems.

This warning indicates that the cluster's routine health check has reported an
unhealthy condition for five minutes. It is a cluster-level signal, rather than
an error for a particular source or sink, and does not necessarily mean that all
workloads are unavailable.

Cloud support handles investigation of the failed check and any required service
repair. Your first step is to check whether your workloads are affected.

## Assess the impact

1. Open the affected cluster in the [Cloud dashboard](https://cloud.risingwave.com/dashboard)
   and select **Metrics**. Choose a time range starting before the alert.

2. Compare these charts with the cluster's normal baseline:

   | Metrics tab | Chart | Look for |
   | - | - | - |
   | **Streaming** | **Streaming Freshness** | Increasing delay in streaming results. |
   | **Streaming** | **Source Rows Processed** | An interruption while upstream data is available. |
   | **Query** | **Query Local Average Latency** and **Query Distributed Average Latency** | Higher latency alongside application errors or timeouts. |
   | **Query** | **Queries Per Second** | An unexpected decrease accompanied by failed application requests. |

3. Check application-side connection and query errors and whether downstream
   results are fresh. Note whether the impact is widespread or limited to
   particular workloads.

If metrics and application behavior remain normal, no customer-side mitigation
is needed. Continue monitoring. Contact support if the alert persists or recurs,
or if you cannot determine the impact.

## Protect connected systems

If workloads are degraded:

* Preserve upstream data retention and buffer incoming work where supported.
* Allow for delayed downstream results, defer non-critical dependent work, or
  use an application fallback if one is available.
* Use your application's retry policy for transient connection or query failures.

Do not delete or recreate resources in response to this health check alone.
Coordinate cluster restarts, **Pause**/**Resume**, and resource changes with
Cloud support.

## Investigation and support

For production impact or a persistent failed check, contact the
[RisingWave Cloud support team](mailto:cloud-support@risingwave-labs.com).
Include the cluster and region, alert time, application errors, affected
workloads, metric screenshots, recent changes, and protective actions already
taken. Support can inspect the health-check results and service logs.

## Confirm recovery

Confirm that queries and streaming workloads return to their normal behavior,
upstream backlogs catch up, and downstream results are fresh. A cleared alert
can also result from missing monitoring data. If business impact continues or
the alert recurs, share the new time window with support.

## Related topics

* [Check status and metrics of clusters](https://docs.risingwave.com/cloud/check-status-and-metrics.md)
* [Monitor messages and alerts](https://docs.risingwave.com/cloud/monitor-messages-and-alerts.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.