pgexporter

Appearance

Custom Metrics ​

Related Documentation

See also: pgexporter Metrics for built-in metrics documentation.

Custom metrics are either defined in a single YAML file, or inside multiple YAML files inside a single directory.

NOTE

Examples of custom YAML files can be found here.

The overall schema for the custom metrics are:

YAML = {
  version : number,
  metrics : Metric[]
};

Metric = {
  queries : Query[],
  tag : string,
  collector : string,
  sort : "name" (default) | "data",
  server : "both" (default) | "primary" | "replica",
};

Query = {
  query : string,
  columns : Column[],
  version ?: number
};

Column = Label | Counter | Gauge;

Label = {
  name: string,
  type: "label"
};

Counter = {
  name ?: string,
  type: "counter",
  description: string
};

Gauge = {
  name ?: string,
  type: "gauge",
  description: string
};

Custom metrics have to be defined in yaml files. There may be a single file or multiple of them (within a single directory).

TIP

Some examples of user-defined metrics can be found here and can be used as a reference while going through this guide.

The structure of these custom yaml files can be approached in a top-down manner as follows.

Top Level ​

At the top-most level of the yaml file, there are two keys version and metrics:

yml
version: ...
metrics: ...
  • version specifies a default minimum PostgreSQL support version for the queries (more on this below).
  • metrics is a list of user defined metrics.

Metrics ​

The metrics key is a list of metrics and has the following structure:

yml
...
metrics:

  - queries: ...
    tag: ...
    collector: ...
    sort: ...
    server: ...

  - queries: ...
    tag: ...
    collector: ...
    sort: ...
    server: ...

  - queries: ...
    tag: ...
    collector: ...
    sort: ...
    server: ...
  ...
  • queries are a list of query alternatives (more on this below).
  • tag: This specifies the tag of the metics (more on this below).
  • collector: This specifies a collector name for the metric. This allows you to disable a collector by using the -C flag (see here for details).
  • sort (optional): This specifies how the output of each metric will be sorted. Currently there are two supported values:
    • name (default): This will sort the output according to server's name.
    • data: This will sort the output according to the data of the first column of the SQL query that will be run on the server (more on this below).
  • server (optional): This specifies on which type of server the metric should query. There are three possible values to this:
    • primary: This means that this metric is only for primary servers.
    • replica: This means that this metric is only for replica servers.
    • both (default): This means that this metric is for both types of servers.

Queries ​

Each queries key is an object of the following structure:

yml
- queries:
  - query: ...
    columns: ...
    version: ...
  - query: ...
    columns: ...
    version: ...
    ...
  • query: Query String for the query alternative (explained below).
  • version (optional): Minimum PostgreSQL version required to run the query (explained below). If this value is not provided, then the default is taken from the top level version.
  • columns: List of columns (explained below)

There are multiple queries for each metrics. They exist because not all queries are supported across all versions of PostgreSQL. The solution is to provide a query, as well as the minimum version of PostgreSQL it will run on.

Thus queries contains multiple entries, each containing a query and a minimum PostgreSQL version required to run it.

Query Selection Based On Server Version ​

Depending on the version of the server, a suitable query is picked. It is picked according to the following rule:

If your server has a version v, then it will select the query with the maximum value of version that it can find which is also less than or equal to v.

For example:

txt
|Server mininum supported PostgreSQL version|Query|
|-------------------------------------------|-----|
|                  10                       |  Q1 |
|                  12                       |  Q2 |
|                  14                       |  Q3 |
|                  16                       |  Q4 |
|                  18                       |  Q5 |
|                  20                       |  Q6 |
|                  22                       |  Q7 |
txt
Server is v9: Not Supported (No query is sent)
Server is v10: Q1 sent
Server is v11: Q1 sent
Server is v12: Q2 sent
Server is v13: Q2 sent
Server is v14: Q3 sent
Server is v15: Q3 sent
Server is v16: Q4 sent
Server is v17: Q4 sent
Server is v18: Q5 sent
Server is v19: Q5 sent
Server is v21: Q6 sent
Server is v25: Q7 sent

Columns ​

Labels, Gauges, Counters and Histograms are the types of Prometheus metrics currently supported by pgexporter. Histogram metric type is a bit different from the rest:

Labels, Gauges and Counters ​

For labels, gauges and counters, columns contains a list of columns, which can be either one of the following:

yml
- name: ...
  type: label

or,

yml
- name: ...
  type: gauge
  description: ...

or,

yml
- name: ...
  type: counter
  description: ...

Gauges or Counters do not strictly need names. It will, by default, take the name from the tag of the metric like pgexporter_tag. However, any names provided will be appended to the tag like: pgexporter_tag_name, and hence names should be provided when dealing with multiple metrics derived from a single query.

NOTE

Names are required for distinction when the same metric has more than one counter/gauge as if not provided, all of these will have the same metric name.

Histograms ​

How to Use Custom Metrics ​

Suppose the user defines their metrics in custom.yml, having the path /location/to/custom.yml. To run pgepxorter with the metrics defined in this file:

sh
$ pgexporter -c pgexporter.conf -u pgexporter_users.conf -Y /location/to/custom.yml

or, if all the custom yaml files are kept in a directory having path /location/to/custom/dir, pgexporter is run as:

sh
$ pgexporter -c pgexporter.conf -u pgexporter_users.conf -Y /location/to/custom/dir

Examples can be refered from here.