Menu
Amazon Redshift
Management Guide (API Version 2012-12-01)

Viewing Cluster Performance Data

Cluster metrics in Amazon Redshift enable the following common performance use cases:

  • Determine if cluster metrics are abnormal over a specified time range and, if so, identify the queries responsible for the performance hit.

  • Check if historical or current queries are impacting cluster performance. If you identify a problematic query, you can view details about it including the cluster performance during the query's execution, information which may assist you in diagnosing why the query was slow, and what can be done to improve its performance.

The default cluster view shows all nodes graphed together, an Average statistic, and data for the last hour. You can change this view as needed. Some metrics, such as HealthStatus, are only applicable for the leader node while others, such as WriteOps, are only applicable for compute nodes. Switching the node display mode will reset all filters.

To view cluster performance data

  1. Sign in to the AWS Management Console and open the Amazon Redshift console at https://console.aws.amazon.com/redshift/.

  2. In the left navigation, click Clusters.

  3. In the Cluster list, click the magnifying glass icon beside the cluster for which you want to view performance data.

  4. Select the Performance tab.

    By default, the performance view displays cluster performance over the past hour. If you need to fine tune the view you have filters that you can use as described in the following table.

    To...Use this filter...
    Change the time range for which data is displayedSelect a time range from the Time Range drop down. By default, the last hour is shown.
    Change the period for which data is displayed

    Select a period from the Period drop down. By default, a 5 minute period is shown. Use a period smaller than 5 minutes if you need more detail when investigating a metric (drilling in) and displaying metrics over a small time period, for example 10 minutes. Similarly, use a period greater than 5 minutes when viewing metrics over a large period of time, for example, days.

    Change the statistic that is displayed for metrics

    Select a statistic from the Statistic drop down. By default, the Average statistic is used.

    Change what metrics are shown, all or a specific metric

    Select a metrics from the Metrics drop down. By default, all metrics are shown.

    Change whether node metrics are displayed separately or together on the same graph

    Click the Nodes button. By default, node data for a given metric is shown on a combined graph. If you choose to display node data on separate graphs, you can additionally show or hide individual nodes.

Cluster Metrics: Examples

The following example shows CPUUtilization and NetworkReceiveThroughput metrics for a single node cluster. In this case the graphs for cluster metrics show one line marked as Shared since the leader and compute node are combined. The example shows that multiple queries were run in the time period shown. On the Queries graph the cursor is positioned over the query running at the peak values of the two metrics and the Query ID is displayed on the right. You could then click the Query ID to find out more about the query running.

The following example shows the NetworkReceiveThroughput for a cluster with two nodes. It shows a line for the leader and two compute nodes. Note that the leader node metrics is flat and is not of interest since data is only loaded on the compute nodes. The example shows that one long query ran in the time period shown. On the Queries graph the cursor is positioned over the long running query and the Query ID is displayed on the right. You could then click the Query ID to find out more about the query running. The NetworkReceiveThroughput value is displayed during the query execution.