View database performance

OCP supports viewing performance monitoring data in a cluster from different dimensions. You can view the data through the following two methods:

• Method 1: Log in to OCP, go to the cluster overview page, and click Performance Monitoring in the left navigation pane.

• Method 2: Log in to OCP, click Performance in the left navigation pane. The system goes to the OceanBase Cluster tab by default.

Prerequisites

• When viewing cluster database performance monitoring data using Method 1, ensure the currently logged-in OCP user has the following permissions:

  • Cluster Read-Only resource permission.
  • Cluster Performance menu permission.

• When viewing cluster database performance monitoring data using Method 2, ensure the currently logged-in OCP user has the following permissions:

  • Any one of the Cluster Read-Only, Tenant Read-Only, or Host Read-Only resource permissions.
  • Performance Monitoring All menu permission.

Version limitations

• For the SQL Response Time monitoring metric:

  • The sub-metrics 99-all and 95-all apply to OceanBase versions [V4.2.1.7, V4.2.2.0).
  • The sub-metrics commit and rollback apply to OceanBase versions [V4.2.5.3, V4.3.0.0), [V4.3.5.2, +∞).

• For the QPS monitoring metric, the sub-metrics commit and rollback apply to OceanBase versions [V4.2.5.3, V4.3.0.0), [V4.3.5.2, +∞).

• For the Transaction Log Time monitoring metric:

  • The sub-metric write_disk applies to OceanBase versions [-∞, V4.1.0.3).
  • The sub-metric clog_write applies to OceanBase versions [V4.1.0.3, +∞).

• For the Physical IO Count, Physical IO Time, and Physical IO Throughput monitoring metrics, the sub-metrics clog_read and clog_write apply to OceanBase versions [V4.1.0.3, +∞).

Procedure

The following procedure uses Method 2 as an example.

1. Log in to OCP.

2. Select Performance Monitoring from the left navigation pane.

3. The system enters the OceanBase Cluster tab by default.

4. In the Data Filtering section, configure the monitoring data filter conditions.

   The filter conditions are described in the following table:

   Filter Condition	Description
   Refresh	Refreshes the monitoring data. Automatic refresh is supported and is disabled by default. When automatic refresh is enabled, you can configure the refresh frequency.
   Time Range (also known as Monitoring Window)	Not displayed in automatic refresh mode. Select a time range to return data within that period.
   Cluster	Select a cluster to monitor.
   Zone	Monitors all zones by default.
   OBServer	Monitors all OBServer nodes by default.

5. The system enters the Database Performance tab by default, where you can view the cluster's database performance monitoring data.

   • OCP supports viewing database performance monitoring data from four dimensions: cluster, tenant, OBServer, and unit. Data is displayed by default at the cluster level. The monitoring metrics supported vary by dimension.
   • The system displays all monitoring items for the current dimension by default. If you do not need to pay attention to a certain monitoring item, you can uncheck it in the Selected Chart panel on the right side of the area.
   • All monitoring items support displaying the maximum value (max), average value (avg), and current value (current) of each sub-metric.
   • For the specific meaning and source of values for all monitoring items, refer to Overview of Monitoring Metrics.

   The following features help you better view and analyze performance monitoring data in trend charts. You can understand them in combination with the images and the following text.

   1. View and hide data for other sub-metrics: Hover the cursor over the metric area and scroll the mouse wheel up or down to view data for all sub-metrics displayed in the trend chart. Click another sub-metric identifier to gray it out, and that metric's information will not be displayed in the trend chart.

   2. Hover the cursor over a point in time in the trend chart to view detailed data for that moment. Drag the cursor to select an area, and the system will zoom in to show the trend information for that area. You can view the detailed trend of the metric during this period. Double-click the trend chart to restore it.

   3. Zoom in on a trend chart: Click the zoom icon in the upper-right corner of the trend chart. Functions ① and ② shown above also apply to the zoomed-in trend chart.

   4. Edit a monitoring metric: Click the edit icon in the upper-right corner of the trend chart to choose which sub-metrics to display.

   5. Drill down to view monitoring data at different dimensions: On the Cluster tab, if you find abnormal monitoring data for a cluster, such as excessively high SQL response time, you can drill down to analyze and locate the server causing the abnormal response time. Click the Drill-down Analysis icon in the upper-right corner of the trend chart. The supported drill-down dimensions are: Cluster > Tenant > OBServer.

      Functions ① and ② shown above also apply to the trend chart on the drill-down page. Additionally, you can:

      1. Reset filter conditions on the drill-down page. If not set, they will automatically inherit the settings from the parent level.

      2. If the metric has sub-metrics, you can select a sub-metric to view; or click ① or ② as shown in the figure to drill down to different dimensions for further analysis.

      3. If the metric involves tenant or host performance, the trend chart provides direct links to the Monitoring page of the relevant tenant or host.

         • To go to the tenant Performance Monitoring page: Hover the mouse over the Go to Tenant Performance Monitoring link in the upper-right corner of the trend chart, select the corresponding tenant, and click.

         • To go to the host Performance Monitoring page: Hover the mouse over the Go to Host Performance Monitoring link in the upper-right corner of the trend chart, select the corresponding host, and click.

6. (Optional) Configure the monitoring collection interval.

   Currently, OCP collects second-level monitoring data by default every 1 second. This collection scale places high demands on both OCP Server performance and MonitorDB storage performance. If your requirements for monitoring data are not stringent, you can modify the value of ocp.metric.collect.interval.second in the config_properties file to configure a different monitoring collection interval, thereby reducing resource consumption and increasing the overall management scale.

   Note

   Currently, only the values 1, 5, 10, and 15 are supported for configuration. If you enter a value other than those listed above, the system will automatically adapt it:

   • Values less than 1 are uniformly set to 1.
   • Values in the [1,5) range are downward-compatible with 1.
   • Values in the [5,10) range are downward-compatible with 5.
   • Values in the [10,15) range are downward-compatible with 10.
   • Other values are uniformly set to 15.

   After the configuration is modified, OCP will immediately start collecting data at the new interval. The display interval for monitoring data defaults to three times the collection interval, except when the collection interval is 1 second. For example, if the collection interval is 5 seconds, the display interval for querying second-level monitoring data will be 15 seconds.