Updated on 2024-12-18 GMT+08:00

Table Diagnosis

GaussDB(DWS) provides statistics and diagnostic tools for you to learn table status, including:

  • Skew Rate: monitors and analyzes uneven data distribution in a cluster, and displays information about the 50 largest tables whose skew rate is higher than 5%.
  • Dirty Page Rate: monitors and analyzes dirty pages in a cluster, and displays information about the 50 largest tables whose dirty page rate is higher than 50%.

Skew Rate

Improper distribution columns can cause severe skew during operator computing or data spill to disk. The workloads will be unevenly distributed on DNs, resulting in high disk usage on individual DNs and affecting their performance. After identifying tables with a high skew rate and a relatively large size, you can reselect distribution columns for these tables to have their data redistributed. For details, see How Do I Change Distribution Columns?

Procedure

  1. Log in to the GaussDB(DWS) console.
  2. Choose Dedicated Clusters > Clusters and locate the cluster to be monitored.
  3. In the Operation column of the target cluster, click Monitoring Panel.
  4. In the navigation tree on the left, choose Utilities > Table Diagnosis and click the Skew Rate tab. The tables that meet the statistics collection conditions in the cluster are displayed.

Dirty Page Rate

DML operations on tables may generate dirty data, which unnecessarily occupies cluster storage. You can identify tables with a high dirty page rate and a relatively large size, and handle them accordingly. For more information, see Solution to High Disk Usage and Cluster Read-Only.

Procedure

  1. Log in to the GaussDB(DWS) console.
  2. Choose Dedicated Clusters > Clusters and locate the cluster to be monitored.
  3. In the Operation column of the target cluster, click Monitoring Panel.
  4. In the navigation tree on the left, choose Utilities > Table Diagnosis and click the Dirty Page Rate tab. The tables that meet the statistics collection conditions in the cluster are displayed.