Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
497 lines (379 sloc) 23.3 KB

grok_exporter Configuration

This page describes the grok_exporter configuration file. The configuration is in YAML format. An example configuration can be found in example/config.yml. The path to the configuration file is passed as a command line parameter when starting grok_exporter:

grok_exporter -config ./example/config.yml

Overall Structure

The grok_exporter configuration file consists of five main sections:

    # Config version
    # How to read log lines (file or stdin).
    # Available Grok patterns.
    # How to map Grok fields to Prometheus metrics.
    # How to expose the metrics via HTTP(S).

The following shows the configuration options for each of these sections.

Global Section

The global section is as follows:

    config_version: 2
    retention_check_interval: 53s

The config_version specifies the version of the config file format. Specifying the config_version is mandatory, it has to be included in every configuration file. The current config_version is 2.

The config file format is versioned independently of the grok_exporter program. When a new version of grok_exporter keeps using the same config file, the config_version will remain the same.

The following table shows which grok_exporter version uses which config_version:

grok_exporter config_version
≤ 0.1.4 1 (see
0.2.X 2 (current version)

The retention_check_interval is the interval at which grok_exporter checks for expired metrics. By default, metrics don't expire so this is relevant only if retention is configured explicitly with a metric. The retention_check_interval is optional, the value defaults to 53s. The default value is reasonable for production and should not be changed. This property is intended to be used in tests, where you might not want to wait 53 seconds until an expired metric is cleaned up. The format is described in How to Configure Durations below.

Input Section

We currently support three input types: file, stdin, and webhook. The following three sections describe the input types respectively:

File Input Type

The configuration for the file input type is as follows:

    type: file
    path: /var/log/sample.log
    readall: false
    fail_on_missing_logfile: true
    poll_interval_seconds: 5 # should not be needed in most cases, see below

The readall flag defines if grok_exporter starts reading from the beginning or the end of the file. True means we read the whole file, false means we start at the end of the file and read only new lines. True is good for debugging, because we process all available log lines. False is good for production, because we avoid to process lines multiple times when grok_exporter is restarted. The default value for readall is false.

If fail_on_missing_logfile is true, grok_exporter will not start if the path is not found. This is the default value, and it should be used in most cases because a missing logfile is likely a configuration error. However, in some scenarios you might want grok_exporter to start successfully even if the logfile is not found, because you know the file will be created later. In that case, set fail_on_missing_logfile: false.

On poll_interval_seconds: You probably don't need this. The internal implementation of grok_exporter's file input is based on the operating system's file system notification mechanism, which is inotify on Linux, kevent on BSD (or macOS), and ReadDirectoryChangesW on Windows. These tools will inform grok_exporter as soon as a new log line is written to the log file, and let grok_exporter sleep as long as the log file doesn't change. There is no need for configuring a poll interval. However, there is one combination where the above notifications don't work: If the logging application keeps the logfile open and the underlying file system is NTFS (see #17). For this specific case you can configure a poll_interval_seconds. This will disable file system notifications and instead check the log file periodically. The poll_interval_seconds option was introduced with release 0.2.2.

Stdin Input Type

The configuration for the stdin input type does not have any additional parameters:

    type: stdin

This is useful if you want to pipe log data to the grok_exporter command, for example if you want to monitor the output of journalctl:

journalctl -f | grok_exporter -config config.yml

Note that grok_exporter terminates as soon as it finishes reading from stdin. That means, if we run cat sample.log | grok_exporter -config config.yml, the exporter will terminate as soon as sample.log is processed, and we will not be able to access the result via HTTP(S) after that. Always use a command that keeps the output open (like tail -f) when testing the grok_exporter with the stdin input.

Webhook Input Type

The grok_exporter is capable of receive log entries from webhook sources. It supports webhook reception in various formats... plain-text or JSON, single entries or bulk entries.

The following input configuration example which demonstrates how to configure grok_exporter to receive HTTP webhooks from the Logstash HTTP Output Plugin configured in json_batch mode, which allows the transmission of multiple json log entries in a single webhook.


    type: webhook

    # HTTP Path to POST the webhook
    # Default is `/webhook`
    webhook_path: /webhook

    # HTTP Body POST Format
    # text_single: Webhook POST body is a single plain text log entry
    # text_bulk: Webhook POST body contains multiple plain text log entries
    #   separated by webhook_text_bulk_separator (default: \n\n)
    # json_single: Webhook POST body is a single json log entry.  Log entry
    #   text is selected from the value of a json key determined by
    #   webhook_json_selector.
    # json_bulk: Webhook POST body contains multiple json log entries.  The
    #   POST body envelope must be a json array "[ <entry>, <entry> ]".  Log
    #   entry text is selected from the value of a json key determined by
    #   webhook_json_selector.
    # Default is `text_single`
    webhook_format: json_bulk

    # JSON Path Selector
    # Within an json log entry, text is selected from the value of this json selector
    #   Example ""
    # Default is `.message`
    webhook_json_selector: .message

    # Bulk Text Separator
    # Separator for text_bulk log entries
    # Default is `\n\n`
    webhook_text_bulk_separator: "\n\n"

This configuration example may be found in the examples directory here.

Grok Section

The grok section configures the available Grok patterns. An example configuration is as follows:

    patterns_dir: ./logstash-patterns-core/patterns
    - 'EXIM_MESSAGE [a-zA-Z ]*'

In most cases, we will have a directory containing the Grok pattern files. Grok's default pattern directory is included in the grok_exporter release. The path to this directory is configured with patterns_dir.

There are two ways to define additional Grok patterns:

  1. Create a custom pattern file and store it in the patterns_dir directory.
  2. Add pattern definitions directly to the grok_exporter configuration. This can be done via the additional_patterns configuration. It takes a list of pattern definitions. The pattern definitions have the same format as the lines in the Grok pattern files.

Grok patterns are simply key/value pairs: The key is the pattern name, and the value is a Grok macro defining a regular expression. There is a lot of documentation available on Grok patterns: The logstash-patterns-core repository contains pre-defined patterns, the Grok documentation shows how patterns are defined, and there are online pattern builders available here: and here:

At least one of patterns_dir or additional_patterns is required: If patterns_dir is missing all patterns must be defined directly in the additional_patterns config. If additional_patterns is missing all patterns must be defined in the patterns_dir.

Metrics Section

Metric Types Overview

The metrics section contains a list of metric definitions, specifying how log lines are mapped to Prometheus metrics. Four metric types are supported:

Example Log Lines

To exemplify the different metrics configurations, we use the following example log lines:

30.07.2016 14:37:03 alice 1.5
30.07.2016 14:37:33 alice 2.5
30.07.2016 14:43:02 bob 2.5
30.07.2016 14:45:59 alice 2.5

Each line consists of a date, time, user, and a number. Using Grok's default patterns, we can create a simple Grok expression matching these lines:



One of the main features of Prometheus is its multi-dimensional data model: A Prometheus metric can be further partitioned using labels.

In order to define a label, you need to first define a Grok field name in the match: pattern, and second add label template under labels:.

  1. Define Grok field names. In Grok, each field, like %{USER}, can be given a name, like %{USER:user}. The name user can then be used in label templates.
  2. Define label templates. Each metric type supports labels, which is a map of name/template pairs. The name will be used in Prometheus as the label name. The template is a Go template that may contain references to Grok fields, like {{.user}}.

Example: In order to define a label user for the example log lines above, use the following fragment:

match: '%{DATE} %{TIME} %{USER:user} %{NUMBER:val}'
    user: '{{.user}}'

The match stores whatever matches the %{USER} pattern under the Grok field name user. The label defines a Prometheus label user with the value of the Grok field user as its content.

This simple example shows a one-to-one mapping of a Grok field to a Prometheus label. However, the label definition is pretty flexible: You can combine multiple Grok fields in one label, and you can define constant labels that don't use Grok fields at all.

Label Template Functions

Label values are defined as Go templates. As of v0.2.6, grok_exporter supports the following template functions: gsub, add, subtract, multiply, divide.

For example, let's assume we have the match from above:

match: '%{DATE} %{TIME} %{USER:user} %{NUMBER:val}'

We apply this pattern to the first log line of our example:

30.07.2016 14:37:03 alice 1.5

Now the Grok field user has the value alice, and the Grok field val has the value 1.5. The following example show how to use these fields as label values using the Go template language:

  • '{{.user}}' -> alice
  • 'user {{.user}} with number {{.val}}.' -> user alice with number 1.5.
  • '{{gsub .user "ali" "beatri"}}' -> beatrice
  • '{{multiply .val 1000}}' -> 1500
  • '{{if eq .user "alice"}}1{{else}}0{{end}}' -> 1

The syntax of the gsub function is {{gsub input pattern replacement}}. The pattern and replacement are is similar to Elastic's mutate filter's gsub (derived from Ruby's String.gsub()), except that you need to double-escape backslashes (\\ instead of \). A more complex example (including capture groups) can be found in this comment.

The arithmetic functions add, subtract, multiply, and divide are straightforward. These functions may not be useful for label values, but they can be useful as the value: in gauge, histogram, or summary metrics. For example, they could be used to convert milliseconds to seconds.

Conditionals like '{{if eq .user "alice"}}1{{else}}0{{end}} are described in the Go template documentation. For example, they can be used to define boolean metrics, i.e. gauge metrics with a value of 1 or 0. Another example can be found in this comment.

Expiring Old Labels

By default, metrics are kept forever. However, sometimes you might want metrics with old labels to expire. There are two ways to do this in grok_exporter:


As of version 0.2.2, grok_exporter supports delete_match and delete_labels configuration:

delete_match: '%{DATE} %{TIME} %{USER:user} logged out'
    user: '{{.user}}'

Without delete_match and delete_labels, all labels are kept forever (until grok_exporter is restarted). However, it might sometimes be desirable to explicitly remove metrics with specific labels. For example, if a service shuts down, it might be desirable to remove metrics labeled with that service name.

Using delete_match you can define a regular expression that will trigger removal of metrics. For example, delete_match could match a shutdown message in a log file.

Using delete_labels you can restrict which labels are deleted if a line matches delete_match. If no delete_labels are specified, all labels for the given metric are deleted. If delete_labels are specified, only those metrics are deleted where the label values are equal to the delete label values.


As of version 0.2.3, grok_exporter supports retention configuration for metrics:

    - type: ...
      name: retention_example
      help: ...
      match: ...
      retention: 2h30m

The example above means that if label values for the metrics named retention_example have not been observed for 2 hours and 30 minutes, the retention_example metrics with these label values will be removed. For the format of the retention value, see How to Configure Durations below. Note that grok_exporter checks the retention every 53 seconds by default, so it may take 53 seconds until the metric is actually removed after the retention time is reached, see retention_check_interval above.

Counter Metric Type

The counter metric counts the number of matching log lines.

    - type: counter
      name: grok_example_lines_total
      help: Example counter metric with labels.
      match: '%{DATE} %{TIME} %{USER:user} %{NUMBER}'
          user: '{{.user}}'

The configuration is as follows:

  • type is counter.
  • name is the name of the metric. Metric names are described in the [Prometheus data model documentation].
  • help is a comment describing the metric.
  • match is the Grok expression. See the Grok documentation for more info.
  • labels is an optional map of name/template pairs, as described above.

Output for the example log lines above:

# HELP grok_example_lines_total Example counter metric with labels.
# TYPE grok_example_lines_total counter
grok_example_lines_total{user="alice"} 3
grok_example_lines_total{user="bob"} 1

Gauge Metric Type

The gauge metric is used to monitor values that are logged with each matching log line.

    - type: gauge
      name: grok_example_values
      help: Example gauge metric with labels.
      match: '%{DATE} %{TIME} %{USER:user} %{NUMBER:val}'
      value: '{{.val}}'
      cumulative: false
          user: '{{.user}}'

The configuration is as follows:

  • type is gauge.
  • name, help, match, and labels have the same meaning as for counter metrics.
  • value is a Go template for the value to be monitored. The template must evaluate to a valid number. The template may use to Grok fields from the match patterns, like the label templates described above.
  • cumulative is optional. By default, the last observed value is measured. With cumulative: true, the sum of all observed values is measured.

Output for the example log lines above::

# HELP grok_example_values Example gauge metric with labels.
# TYPE grok_example_values gauge
grok_example_values{user="alice"} 6.5
grok_example_values{user="bob"} 2.5

Histogram Metric Type

Like gauge metrics, the histogram metric monitors values that are logged with each matching log line. However, instead of just summing up the values, histograms count the observed values in configurable buckets.

    - type: histogram
      name: grok_example_values
      help: Example histogram metric with labels.
      match: '%{DATE} %{TIME} %{USER:user} %{NUMBER:val}'
      value: '{{.val}}'
      buckets: [1, 2, 3]
          user: '{{.user}}'

The configuration is as follows:

  • type is histogram.
  • name, help, match, labels, and value have the same meaning as for gauge metrics.
  • buckets configure the categories to be observed. In the example, we have 4 buckets: One for values < 1, one for values < 2, one for values < 3, and one for all values (i.e. < infinity). Buckets are optional. The default buckets are [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10], which is useful for HTTP response times in seconds.

Output for the example log lines above::

# HELP grok_example_values Example histogram metric with labels.
# TYPE grok_example_values histogram
grok_example_values_bucket{user="alice",le="1"} 0
grok_example_values_bucket{user="alice",le="2"} 1
grok_example_values_bucket{user="alice",le="3"} 3
grok_example_values_bucket{user="alice",le="+Inf"} 3
grok_example_values_sum{user="alice"} 6.5
grok_example_values_count{user="alice"} 3
grok_example_values_bucket{user="bob",le="1"} 0
grok_example_values_bucket{user="bob",le="2"} 0
grok_example_values_bucket{user="bob",le="3"} 1
grok_example_values_bucket{user="bob",le="+Inf"} 1
grok_example_values_sum{user="bob"} 2.5
grok_example_values_count{user="bob"} 1

Summary Metric Type

Like gauge and histogram metrics, the summary metric monitors values that are logged with each matching log line. Summaries measure configurable φ quantiles, like the median (φ=0.5) or the 95% quantile (φ=0.95). See histograms and summaries for more info.

   - type: summary
      name: grok_example_values
      help: Summary metric with labels.
      match: '%{DATE} %{TIME} %{USER:user} %{NUMBER:val}'
      value: '{{.val}}'
      quantiles: {0.5: 0.05, 0.9: 0.01, 0.99: 0.001}
          user: '{{.user}}'

The configuration is as follows:

  • type is summary.
  • name, help, match, labels, and value have the same meaning as for gauge metrics.
  • quantiles is a list of quantiles to be observed. grok_exporter does not provide exact values for the quantiles, but only estimations. For each quantile, you also specify an uncertainty that is tolerated for the estimation. In the example, we measure the median (0.5 quantile) with uncertainty 5%, the 90% quantile with uncertainty 1%, and the 99% quantile with uncertainty 0.1%. quantiles is optional, the default value is {0.5: 0.05, 0.9: 0.01, 0.99: 0.001}.

Summaries represent a sliding time window of 10 minutes, i.e. if you observe a 0.5 quantile (median) of x, the value x represents the median within the last 10 minutes. The time window is moved forward every 2 minutes.

Output for the example log lines above::

# HELP grok_example_values Example summary metric with labels.
# TYPE grok_example_values summary
grok_example_values{user="alice",quantile="0.5"} 2.5
grok_example_values{user="alice",quantile="0.9"} 2.5
grok_example_values{user="alice",quantile="0.99"} 2.5
grok_example_values_sum{user="alice"} 6.5
grok_example_values_count{user="alice"} 3
grok_example_values{user="bob",quantile="0.5"} 2.5
grok_example_values{user="bob",quantile="0.9"} 2.5
grok_example_values{user="bob",quantile="0.99"} 2.5
grok_example_values_sum{user="bob"} 2.5
grok_example_values_count{user="bob"} 1

Server Section

The server section configures the HTTP(S) server for exposing the metrics:

    protocol: https
    host: localhost
    port: 9144
    path: /metrics
    cert: /path/to/cert
    key: /path/to/key
  • protocol can be http or https. Default is http.
  • host can be a hostname or an IP address. If host is specified, grok_exporter will listen on the network interface with the given address. If host is omitted, grok_exporter will listen on all available network interfaces. If host is set to [::], grok_exporter will listen on all IPV6 addresses.
  • port is the TCP port to be used. Default is 9144.
  • path is the path where the metrics are exposed. Default is /metrics, i.e. by default metrics will be exported on http://localhost:9144/metrics.
  • cert is the path to the SSL certificate file for protocol https. It is optional. If omitted, a hard-coded default certificate will be used.
  • key is the path to the SSL key file for protocol https. It is optional. If omitted, a hard-coded default key will be used.

How to Configure Durations

grok_exporter uses the format from golang's time.ParseDuration() for configuring time intervals. Some examples are:

  • 2h30m: 2 hours and 30 minutes
  • 100ms: 100 milliseconds
  • 1m30s: 1 minute and 30 seconds
  • 5m: 5 minutes
You can’t perform that action at this time.