Inspecting Network Performance

The Network Performance Inspector allows you to examine the latency among the hosts managed by Cloudera Manager. You can use this tool to diagnose latency issues that can significantly affect the performance of workloads such as MapReduce jobs, Spark jobs, and Hive and Impala queries, particularly when using remote storage.

The inspector runs ping commands from each host to all other hosts, and reports the average ping time and packet loss percentage. You can use this information to identify problematic hosts or networking infrastructure issues so that you take corrective action. You can run the inspector on-demand, and it also available when adding a new cluster. You can also run the inspector using the Cloudera Manager API.

Running the Network Performance Inspector From the Cloudera Manager Admin Console

To run the Network Performance Inspector:
  1. Open the Network Performance Inspector from one of the following pages in the Cloudera Manager Admin Console:
    • From the All Hosts page:
      1. Select Hosts > All Hosts.
      2. Click the Inspect Network Performance button to launch the inspector.
    • From the Cluster Status page:
      1. Select Clusters > Cluster Name
      2. In the Status section, click the Hosts link at the top of the list of Cluster Services to open the Hosts page.
      3. Click the Inspect Cluster Network Performance button to launch the inspector.

    The Inspect Network Performance dialog box opens.

  2. Enter the following information:
    • Source Cluster – Select a cluster to inspect from the drop-down list.

      Select Run against another cluster to select additional clusters. An additional drop-down list displays from which you can select an additional cluster to inspect.

    • Ping Timeout – Amount of time, in seconds, after which the inspector reports a failure.
    • Ping Count – Number of times the inspector pings each host.
    • Ping Packet Size – Size, in bytes, of the test packet sent when pinging the hosts.
  3. Click Run.

    The Cluster Performance Inspector command window opens. When the Inspector finishes, click the Show Inspector Results button to open the Network Performance Inspector Results page.

    Network Performance Inspector Result Page

    Select Show Hosts with Issues to display any problems found by the inspector, or select Show All Hosts. The Network Diagnostic Result page displays a table of results. Each row represents a single host. The Target Hosts Summary column summarizes the performance of the host. Click the summary text to view detailed performance statistics from this host to each of the other hosts.

    The inspector summarizes the status of the hosts using three icons:
    • Green – Good network performance.
    • Orange – Concerning network performance. Any host with a ping latency in the range of 1 to 4 milliseconds.
    • Red – Bad network performance. Any target host that is unreachable by hostname (has a 100% packet loss), any host with a ping latency greater than 4 milliseconds, or any target host with a packet loss of 1% or greater.

    Poor performance can result from firewall settings, router configurations, network topology, and other factors. You may need to work with your network administrator to mitigate these issues.

Running the Network Performance Inspector From the Cloudera Manager API

You can invoke the Network Performance Inspector using the Cloudera Manager API and the following endpoints:
  • /cm/commmands/hostsPerfInspector

    Invokes the inspector across an arbitrary set of hosts, (including hosts that are not part of the cluster).

  • /cm/commands/clusterPerfInspector

    Invokes the inspector across the hosts in two clusters.

  • /clusters/cluster Name/commands/perfInspector

    Invokes the inspector across the hosts of a specified cluster.

For more information, see the Cloudera Manager REST API documentation