Chapter 4. PCP Metrics

Metrics are stored in a tree-like structure. For example, all network metrics are stored in a node named network. Each metric may be a single value, or a list of values, known as instances. For example, kernel load has three instances, a 1-minute, 5-minute, and 15-minute average.

For every metric entry, PCP stores both its data and metadata. This includes the metrics description, data type, units, and dimensions. For example, the metadata enables PCP to output multiple metrics with different dimensions.

The value of a counter metric only increases. For example, a count of disk write operations on a specific device only increases. When you query the value of a counter metric, PCP converts this into a rate value by default.

In addition to system metrics such as CPU, memory, kernel, XFS, disk, and network, the following metrics are configured:

MetricDescription

hotproc.*

Basic metrics of key Satellite processes

apache.*

Apache HTTP Server metrics

postgresql.*

Basic PostgreSQL statistics

mmv.fm_rails_*

Satellite metrics

4.1. Identifying Available Metrics

  • To list all metrics available via PCP, enter the following command:

    # pminfo
  • To list all Satellite metrics and their descriptions, enter the following command:

    # foreman-rake telemetry:metrics
  • To list the archived metrics, enter the following command:

    # less /var/log/pcp/pmlogger/$(hostname)/pmlogger.log
  • The pmlogger daemon archives data as it is received, according to its configuration. To confirm the active archive file, enter the following command:

    # pcp | grep logger

    The output includes the file name of the active archive file, for example:

    /var/log/pcp/pmlogger/satellite.example.com/20180814.00.10