Symptoms

  • Partition skew causes high usage of some disks in the cluster and insufficient utilization of disk performance and capacity.
  • Partition skew also results in large amounts of traffic in some nodes and triggers single-node throttling.

Causes

When you create a topic, the number of partitions is not set to a value that is a multiple of the number that is recommended by the system.

Solutions

  1. Log on to the ApsaraMQ for Kafka console. In the left-side navigation pane, click Instances.
  2. In the top navigation bar, select a region and click the name of the instance that you want to manage on the Instances page.
  3. In the left-side navigation pane, click Topics. On the Topics page, click the name of the topic that you want to manage.
  4. In the Configuration Information tab of the Topic Details page, click Increase Partitions. In the panel that appears, click the upward or downward arrow for the Partitions parameter to change the number to a multiple of the recommended number.
    If you cannot increase the number of partitions in a topic that uses cloud storage for business reasons, you can use the partition balancing feature to change the location of the partitions. After the change, new data in the partitions is redirected to a disk that has low usage. After messages in the topic expire, the data on the original disk is deleted and disk usage is reduced. For more information, see Partition balancing and traffic redirection.