Cluster health inspections are used to periodically or temporarily assess the overall health of an OceanBase cluster, producing a structured inspection summary and risk alerts.
When to use
Conduct a comprehensive check-up on core clusters before or after shift handovers.
Verify cluster resources and alert status before critical events such as major business promotions.
When user feedback is received but the issue is unclear, perform a thorough inspection first to narrow down the troubleshooting scope.
Prerequisites
You have configured an OCP connection for the product to which the target cluster belongs. If diagnostic metrics such as slow SQL and top SQL need to be inspected, you must also configure an OAS connection.
The
cluster_inspectionskill is enabled.
Procedure
On the Create Dialog page, enter
@and select the target product and cluster.Enter your inspection request, for example:
Help me inspect this cluster.You can also click the shortcut entry Cluster Health Inspection and then specify the cluster and time range.
Press Enter to send the request and wait for the Agent to complete the analysis.
The Agent will sequentially query multi-dimensional data such as the cluster's basic information, resource usage (CPU/memory/disk), current active alerts, and an overview of slow queries and top SQL.
View the inspection conclusion.
The conclusion output by the Agent typically includes the following sections:
Problem Overview: The core issues identified in this inspection and their impact.
Root Cause Analysis: The possible root cause of the problem, along with supporting evidence such as waiting events, proportion, and involved background processes.
Key Evidence: Cited raw metric data, for example, sampling proportion of Top Background DB Time, tenant load characteristics, I/O bandwidth data, etc.
Solution: Troubleshooting and remediation steps prioritized by order.
What to do next
Expand Tool Output to view the key metrics referenced in the verification conclusion. For more information, see View the tool call process.
Configure inspection to run automatically daily or weekly. For more information, see Create a scheduled task.
If an inspection identifies slow SQL risks, you can drill down for analysis within the same session. For more information, see Slow SQL, TopSQL, and abnormal SQL analysis.