PostgreSQL Server Exporter

See the docs/ directory for user-focused documentation: getting started, Docker images, configuration, database permissions, connecting, and secrets.

Quick Start

# Start an example database
docker run --net=host -it --rm -e POSTGRES_PASSWORD=password postgres
Connect to it
docker run \
  --net=host \
  -e DATA_SOURCE_URI="localhost:5432/postgres?sslmode=disable" \
  -e DATA_SOURCE_USER=postgres \
  -e DATA_SOURCE_PASS=password \
  quay.io/prometheuscommunity/postgres-exporter

Test with:

curl "http://localhost:9187/metrics"

See Getting Started for the other ways to install it, how to point Prometheus at it, and how to supply the password without putting it in an environment variable.

Multi-Target Support

This exporter supports the multi-target pattern. This allows running a single instance of this exporter for multiple postgres targets. Using the multi-target functionality of this exporter is optional and meant for cases where it is impossible to install the exporter as a sidecar, for example SaaS-managed services.

To use the multi-target functionality, send an http request to the endpoint /probe?target=foo:5432 where target is set to the DSN of the postgres instance to scrape metrics from.

To avoid putting sensitive information like username and password in the URL, preconfigured auth modules are supported via the auth_modules section of the config file. auth_modules for DSNs can be used with the /probe endpoint by specifying the ?auth_module=foo http parameter.

Example Prometheus config:

scrape_configs:
  - job_name: 'postgres'
    static_configs:
      - targets:
        - server1:5432
        - server2:5432
    metrics_path: /probe
    params:
      auth_module: [foo]
    relabel_configs:
      - source_labels: [__address__]
        target_label: __param_target
      - source_labels: [__param_target]
        target_label: instance
      - target_label: __address__
        replacement: 127.0.0.1:9116  # The postgres exporter's real hostname:port.

Configuration File

The configuration file controls the behavior of the exporter. It can be set using the --config.file command line flag and defaults to postgres_exporter.yml. Set --config.file="" to run without a configuration file.

auth_modules

Example:

auth_modules:
  foo1: # Set this to any name you want
    type: userpass
    userpass:
      username: first
      password: firstpass
    options:
      # options become key=value parameters of the DSN
      sslmode: disable

Building and running

git clone https://github.com/prometheus-community/postgres_exporter.git cd postgres_exporter make build ./postgres_exporter

To build the Docker image:

make promu promu crossbuild -p linux/amd64 -p linux/armv7 -p linux/arm64 -p linux/ppc64le make docker

This will build the docker image as prometheuscommunity/postgres_exporter:${branch}.

Flags

• help

• [no-]collector.aurora_global_db_status

• [no-]collector.aurora_replica_status

• [no-]collector.database

• [no-]collector.database_wraparound

• [no-]collector.locks

• [no-]collector.long_running_transactions

• [no-]collector.postmaster

• [no-]collector.process_idle

• [no-]collector.replication

• [no-]collector.replication_slots

• [no-]collector.settings

• [no-]collector.stat_activity

• [no-]collector.stat_activity_autovacuum

• [no-]collector.stat_archiver

• [no-]collector.stat_bgwriter

• [no-]collector.stat_checkpointer

• [no-]collector.stat_database

• [no-]collector.stat_progress_vacuum

• [no-]collector.stat_replication

• [no-]collector.stat_statements

• [no-]collector.stat_statements.include_query

• --collector.stat_statements.query_length

• --collector.stat_statements.limit

• --collector.stat_statements.exclude_databases

• --collector.stat_statements.exclude_users

• [no-]collector.stat_user_tables

• [no-]collector.stat_wal_receiver

• [no-]collector.statio_user_indexes

• [no-]collector.statio_user_tables

• [no-]collector.wal

• [no-]collector.xlog_location

• config.file

• web.systemd-socket

• web.listen-address

• web.config.file

• web.telemetry-path

• disable-default-metrics

• auto-discover-databases (DEPRECATED)

• auto-discover-databases.max-concurrency

• extend.query-path (DEPRECATED)

• dumpmaps

• [no-]wrap-large-counters

• constantLabels (DEPRECATED)

• version

• exclude-databases (DEPRECATED)

• include-databases (DEPRECATED)

• log.level

• log.format

Environment Variables

The following environment variables configure the exporter:

• DATA_SOURCE_NAME

• DATA_SOURCE_URI

• DATA_SOURCE_URI_FILE

• DATA_SOURCE_USER

• DATA_SOURCE_USER_FILE

• DATA_SOURCE_PASS

• DATA_SOURCE_PASS_FILE

• PG_EXPORTER_COLLECTION_TIMEOUT

• PG_EXPORTER_WEB_TELEMETRY_PATH

• PG_EXPORTER_DISABLE_DEFAULT_METRICS

• PG_EXPORTER_AUTO_DISCOVER_DATABASES (DEPRECATED)

• PG_EXPORTER_EXTEND_QUERY_PATH

• PG_EXPORTER_CONSTANT_LABELS (DEPRECATED)

• PG_EXPORTER_EXCLUDE_DATABASES (DEPRECATED)

• PG_EXPORTER_INCLUDE_DATABASES (DEPRECATED)

• PG_EXPORTER_METRIC_PREFIX

Settings set by environment variables starting with PG_ will be overwritten by the corresponding CLI flag if given.

Setting the Postgres server's data source name

The PostgreSQL server's data source name must be set via the DATA_SOURCE_NAME environment variable.

For running it locally on a default Debian/Ubuntu install, this will work (transpose to init script as appropriate):

sudo -u postgres DATA_SOURCE_NAME="user=postgres host=/var/run/postgresql/ sslmode=disable" postgres_exporter

Also, you can set a list of sources to scrape different instances from the one exporter setup. Just define a comma separated string.

sudo -u postgres DATA_SOURCE_NAME="port=5432,port=6432" postgres_exporter

See the github.com/lib/pq module for other ways to format the connection string.

Adding new metrics

The exporter will attempt to dynamically export additional metrics if they are added in the future, but they will be marked as "untyped". Additional metric maps can be easily created from Postgres documentation by copying the tables and using the following Python snippet:

x = """tab separated raw text of a documentation table"""
for l in StringIO(x):
    column, ctype, description = l.split('\t')
    print """"{0}" : {{ prometheus.CounterValue, prometheus.NewDesc("pg_stat_database_{0}", "{2}", nil, nil) }}, """.format(column.strip(), ctype, description.strip())

Adding new metrics via a config file (DEPRECATED)

This feature is deprecated in favor of built-in collector functions. For generic SQL database monitoring see the sql_exporter.

The -extend.query-path command-line argument specifies a YAML file containing additional queries to run. Some examples are provided in queries.yaml.

Disabling default metrics

Automatically discover databases (DEPRECATED)

Every discovered database gets its own connection and gets scraped for the collectors that report on a single database (e.g. stat_user_tables, statio_user_tables, statio_user_indexes). Collectors that report on the whole server (e.g. wal, bgwriter, stat_activity) still only run once, against the first configured DSN, since re-running them against another database on the same server would report the exact same rows again. The set of discovered databases is re-evaluated on every scrape, so databases created or dropped after the exporter starts are picked up without a restart. At most --auto-discover-databases.max-concurrency discovered databases are scraped concurrently, so a server with many databases cannot make a single scrape open more connections than it can spare.

In addition, the option --exclude-databases adds the possibily to filter the result from the auto discovery to discard databases you do not need.

If you want to include only subset of databases, you can use option --include-databases. Exporter still makes request to pg_database table, but do scrape from only if database is in include list.

Running as non-superuser

To be able to collect metrics from pg_stat* views as non-superuser in PostgreSQL server versions >= 10 you can grant the pg_monitor or pg_read_all_stats built-in roles to the user. If you need to monitor older PostgreSQL servers, you will have to create functions and views as a superuser, and assign permissions separately to those.

-- To use IF statements, hence to be able to check if the user exists before
-- attempting creation, we need to switch to procedural SQL (PL/pgSQL)
-- instead of standard SQL.
-- More: https://www.postgresql.org/docs/9.3/plpgsql-overview.html
-- To preserve compatibility with <9.0, DO blocks are not used; instead,
-- a function is created and dropped.
CREATE OR REPLACE FUNCTION __tmp_create_user() returns void as $$
BEGIN
  IF NOT EXISTS (
          SELECT                       -- SELECT list can stay empty for this
          FROM   pg_catalog.pg_user
          WHERE  usename = 'postgres_exporter') THEN
    CREATE USER postgres_exporter;
  END IF;
END;
$$ language plpgsql;
SELECT __tmp_create_user();
DROP FUNCTION __tmp_create_user();
ALTER USER postgres_exporter WITH PASSWORD 'password';
ALTER USER postgres_exporter SET SEARCH_PATH TO postgres_exporter,pg_catalog;
-- If deploying as non-superuser (for example in AWS RDS), uncomment the GRANT
-- line below and replace <MASTER_USER> with your root user.
-- GRANT postgres_exporter TO <MASTER_USER>;
GRANT CONNECT ON DATABASE postgres TO postgres_exporter;

Run following command if you use PostgreSQL versions >= 10

GRANT pg_monitor to postgres_exporter;

Run following SQL commands only if you use PostgreSQL versions older than 10. In PostgreSQL, views run with the permissions of the user that created them so they can act as security barriers. Functions need to be created to share this data with the non-superuser. Only creating the views will leave out the most important bits of data.

CREATE SCHEMA IF NOT EXISTS postgres_exporter;
GRANT USAGE ON SCHEMA postgres_exporter TO postgres_exporter;

CREATE OR REPLACE FUNCTION get_pg_stat_activity() RETURNS SETOF pg_stat_activity AS
$$ SELECT * FROM pg_catalog.pg_stat_activity; $$
LANGUAGE sql
VOLATILE
SECURITY DEFINER;
CREATE OR REPLACE VIEW postgres_exporter.pg_stat_activity
AS
  SELECT * from get_pg_stat_activity();
GRANT SELECT ON postgres_exporter.pg_stat_activity TO postgres_exporter;
CREATE OR REPLACE FUNCTION get_pg_stat_replication() RETURNS SETOF pg_stat_replication AS
$$ SELECT * FROM pg_catalog.pg_stat_replication; $$
LANGUAGE sql
VOLATILE
SECURITY DEFINER;
CREATE OR REPLACE VIEW postgres_exporter.pg_stat_replication
AS
  SELECT * FROM get_pg_stat_replication();
GRANT SELECT ON postgres_exporter.pg_stat_replication TO postgres_exporter;
CREATE EXTENSION IF NOT EXISTS pg_stat_statements;
CREATE OR REPLACE FUNCTION get_pg_stat_statements() RETURNS SETOF pg_stat_statements AS
$$ SELECT * FROM public.pg_stat_statements; $$
LANGUAGE sql
VOLATILE
SECURITY DEFINER;
CREATE OR REPLACE VIEW postgres_exporter.pg_stat_statements
AS
  SELECT * FROM get_pg_stat_statements();
GRANT SELECT ON postgres_exporter.pg_stat_statements TO postgres_exporter;

NOTE

Remember to use postgres database name in the connection string:

> DATA_SOURCE_NAME=postgresql://postgres_exporter:password@localhost:5432/postgres?sslmode=disable

Running the tests

# Run the unit tests
make test
Start the test database with docker
docker run -p 5432:5432 -e POSTGRES_DB=circle_test -e POSTGRES_USER=postgres -e POSTGRES_PASSWORD=test -d postgres
Run the integration tests
DATA_SOURCE_NAME='postgresql://postgres:test@localhost:5432/circle_test?sslmode=disable' GOOPTS='-v -tags integration' make test