Table of ContentsView in Frames

Viewing performance data for a time period

You can monitor cluster performance by collecting and viewing data for a specific time period (a sample). You can view data for several objects and instances at a time.

About this task

You can collect more than one data sample at a time. You can collect more than one sample from the same object at the same time.
Note: You cannot collect and view data for an object that has more than 5,000 instances. If an object has more than 5,000 instances, you need to specify the specific instances for which you want data. This applies to all statistics commands, including statistics views.

For more information about the statistics commands, see the man pages.

Steps

  1. Switch to the advanced privilege level by using the set -privilege advanced command.
  2. Use the statistics start command to start collecting data.
    If you do not specify the -sample-id parameter, the command generates a sample identifier for you and defines this sample as the default sample for the CLI session. If you run this command again during the same CLI session and do not specify the -sample-id parameter, the command can overwrite the previous default sample. You are prompted to confirm whether to overwrite the previous default sample.
  3. Optional: Use the statistics stop command to stop collecting data for the sample.
    You can view data from the sample if you do not stop data collection. Stopping data collection gives you a fixed sample. Not stopping data collection gives you the ability to get updated data that you can use to compare against previous queries. The comparison can help you identify performance trends.
  4. Use the statistics show command to view the sample data.

Example: Monitoring NFSv3 performance

The following example shows performance data for the NFSv3 protocol.

The following command starts data collection for a new sample:

cluster1::*> statistics start -object nfsv3 -sample-id nfs_sample

The following command shows data from the sample by specifying counters that show the number of successful read and write requests versus the total number of read and write requests:

cluster1::*> statistics show -sample-id nfs_sample -counter read_total|write_total|read_success|write_success

Object: nfsv3
Instance: vs1
Start-time: 2/11/2013 15:38:29
End-time: 2/11/2013 15:38:41
Cluster: cluster1

    Counter                                           Value
    --------------------------- ---------------------------
    read_success                                      40042
    read_total                                        40042
    write_success                                   1492052
    write_total                                     1492052