OpenMetrics input data format

Use the openmetrics input data format to parse OpenMetrics text format into Telegraf metrics.

The prometheus input plugin uses this parser internally when an endpoint serves OpenMetrics. Use the parser directly with a listener plugin, such as http_listener_v2, to accept pushed OpenMetrics data in the style of a Pushgateway.

Configuration

[[inputs.http_listener_v2]]
  service_address = ":8080"
  paths = ["/metrics"]
  data_format = "openmetrics"

  ## Use the timestamp from the parsed data (default) or the time of
  ## parsing.
  # openmetrics_ignore_timestamp = false

  ## Metric layout to produce (see below).
  # openmetrics_metric_version = 2

openmetrics_ignore_timestamp

If true, the parser discards timestamps included in the source data and assigns the time of parsing instead.

Type: boolean
Default: false

openmetrics_metric_version

Controls how OpenMetrics metrics translate to Telegraf metrics.

Type: integer (1 or 2)
Default: 2

Metric layouts

Given the following OpenMetrics input:

# TYPE go_goroutines gauge
# HELP go_goroutines Number of goroutines that currently exist.
go_goroutines 69
# TYPE process_cpu_seconds counter
# UNIT process_cpu_seconds seconds
# HELP process_cpu_seconds Total user and system CPU time spent in seconds.
process_cpu_seconds_total 4.20072246e+06
# EOF

Version 2 (default)

Each OpenMetrics MetricPoint becomes a Telegraf metric named prometheus. Labels become tags and the field name is based on the OpenMetrics metric name:

prometheus go_goroutines=69
prometheus,unit=seconds process_cpu_seconds=4200722.46

The resulting metrics are sparse, but often easier to process and query for destinations that are more efficient with column-oriented data. To change the prometheus metric name, use the name_override plugin option. To produce multiple metric names, use multiple instances of the input plugin, each with its own name_override.

Histograms use the same format as the histogram aggregator.

Version 1

The OpenMetrics metric-family name becomes the Telegraf metric name, labels become tags, values become fields, and field names are based on the type of the OpenMetrics metric:

go_goroutines gauge=69
process_cpu_seconds,unit=seconds counter=4200722.46

The resulting metrics are dense, which is efficient for destinations with row-oriented data models.