Database Monitoring provides deep visibility into your Postgres databases by exposing query metrics, query samples, explain plans, database states, failovers, and events.

The Agent collects telemetry directly from the database by logging in as a read-only user. Do the following setup to enable Database Monitoring with your Postgres database:

1. Configure database parameters
2. Grant the Agent access to the database
3. Install the Agent
4. Install the Cloud SQL Integration

Before you begin

Supported PostgreSQL versions

10, 11, 12, 13, 14, 15

Supported Agent versions

7.36.1+

Performance impact

The default Agent configuration for Database Monitoring is conservative, but you can adjust settings such as the collection interval and query sampling rate to better suit your needs. For most workloads, the Agent represents less than one percent of query execution time on the database and less than one percent of CPU.

Database Monitoring runs as an integration on top of the base Agent (see benchmarks).

Proxies, load balancers, and connection poolers

The Datadog Agent must connect directly to the host being monitored. For self-hosted databases, 127.0.0.1 or the socket is preferred. The Agent should not connect to the database through a proxy, load balancer, or connection pooler such as pgbouncer. If the Agent connects to different hosts while it is running (as in the case of failover, load balancing, and so on), the Agent calculates the difference in statistics between two hosts, producing inaccurate metrics.

Data security considerations

See Sensitive information for information about what data the Agent collects from your databases and how to ensure it is secure.

Configure Postgres settings

Configure the following parameters in Database flags and then restart the server for the settings to take effect. For more information about these parameters, see the Postgres documentation.

Grant the Agent access

The Datadog Agent requires read-only access to the database server in order to collect statistics and queries.

The following SQL commands should be executed on the primary database server (the writer) in the cluster if Postgres is replicated. Choose a PostgreSQL database on the database server for the Agent to connect to. The Agent can collect telemetry from all databases on the database server regardless of which one it connects to, so a good option is to use the default postgres database. Choose a different database only if you need the Agent to run custom queries against data unique to that database.

Connect to the chosen database as a superuser (or another user with sufficient permissions). For example, if your chosen database is postgres, connect as the postgres user using psql by running:

psql -h mydb.example.com -d postgres -U postgres

Create the datadog user:

CREATE USER datadog WITH password '<PASSWORD>';

Create the following schema in every database:

CREATE SCHEMA datadog;
GRANT USAGE ON SCHEMA datadog TO datadog;
GRANT USAGE ON SCHEMA public TO datadog;
GRANT pg_monitor TO datadog;
CREATE EXTENSION IF NOT EXISTS pg_stat_statements;

Create the function in every database to enable the Agent to collect explain plans.

CREATE OR REPLACE FUNCTION datadog.explain_statement(
   l_query TEXT,
   OUT explain JSON
)
RETURNS SETOF JSON AS
$$
DECLARE
curs REFCURSOR;
plan JSON;

BEGIN
   OPEN curs FOR EXECUTE pg_catalog.concat('EXPLAIN (FORMAT JSON) ', l_query);
   FETCH curs INTO plan;
   CLOSE curs;
   RETURN QUERY SELECT plan;
END;
$$
LANGUAGE 'plpgsql'
RETURNS NULL ON NULL INPUT
SECURITY DEFINER;

Verify

To verify the permissions are correct, run the following commands to confirm the Agent user is able to connect to the database and read the core tables:

psql -h localhost -U datadog postgres -A \
  -c "select * from pg_stat_database limit 1;" \
  && echo -e "\e[0;32mPostgres connection - OK\e[0m" \
  || echo -e "\e[0;31mCannot connect to Postgres\e[0m"
psql -h localhost -U datadog postgres -A \
  -c "select * from pg_stat_activity limit 1;" \
  && echo -e "\e[0;32mPostgres pg_stat_activity read OK\e[0m" \
  || echo -e "\e[0;31mCannot read from pg_stat_activity\e[0m"
psql -h localhost -U datadog postgres -A \
  -c "select * from pg_stat_statements limit 1;" \
  && echo -e "\e[0;32mPostgres pg_stat_statements read OK\e[0m" \
  || echo -e "\e[0;31mCannot read from pg_stat_statements\e[0m"

When it prompts for a password, use the password you entered when you created the datadog user.

Install the Agent

To monitor Cloud SQL hosts, install the Datadog Agent in your infrastructure and configured to connect to each instance remotely. The Agent does not need to run on the database, it only needs to connect to it. For additional Agent installation methods not mentioned here, see the Agent installation instructions.

export DD_API_KEY=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
export DD_AGENT_VERSION=7.36.1

docker run -e "DD_API_KEY=${DD_API_KEY}" \
  -v /var/run/docker.sock:/var/run/docker.sock:ro \
  -l com.datadoghq.ad.check_names='["postgres"]' \
  -l com.datadoghq.ad.init_configs='[{}]' \
  -l com.datadoghq.ad.instances='[{
    "dbm": true,
    "host": "<INSTANCE_ADDRESS>",
    "port": 5432,
    "username": "datadog",
    "password": "<UNIQUEPASSWORD>",
    "gcp": {
      "project_id": "<PROJECT_ID>",
      "instance_id": "<INSTANCE_ID>"
    }
  }]' \
  gcr.io/datadoghq/agent:${DD_AGENT_VERSION}

FROM gcr.io/datadoghq/agent:7.36.1

LABEL "com.datadoghq.ad.check_names"='["postgres"]'
LABEL "com.datadoghq.ad.init_configs"='[{}]'
LABEL "com.datadoghq.ad.instances"='[{"dbm": true, "host": "<INSTANCE_ADDRESS>", "port": 5432,"username": "datadog","password": "<UNIQUEPASSWORD>", "gcp": {"project_id": "<PROJECT_ID>", "instance_id": "<INSTANCE_ID>"}}]'

helm repo add datadog https://helm.datadoghq.com
helm repo update

helm install <RELEASE_NAME> \
  --set 'datadog.apiKey=<DATADOG_API_KEY>' \
  --set 'clusterAgent.enabled=true' \
  --set 'clusterChecksRunner.enabled=true' \
  --set 'clusterAgent.confd.postgres\.yaml=cluster_check: true
init_config:
instances:
  - dbm: true
    host: <INSTANCE_ADDRESS>
    port: 5432
    username: datadog
    password: "<UNIQUEPASSWORD>"
    gcp:
      project_id: "<PROJECT_ID>"
      instance_id: "<INSTANCE_ID>"' \
  datadog/datadog

cluster_check: true  # Make sure to include this flag
init_config:
instances:
  - dbm: true
    host: '<INSTANCE_ADDRESS>'
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    # After adding your project and instance, configure the Datadog GCP integration to pull additional cloud data such as CPU, Memory, etc.
    gcp:
      project_id: '<PROJECT_ID>'
      instance_id: '<INSTANCE_ID>'

apiVersion: v1
kind: Service
metadata:
  name: postgres
  labels:
    tags.datadoghq.com/env: '<ENV>'
    tags.datadoghq.com/service: '<SERVICE>'
  annotations:
    ad.datadoghq.com/service.check_names: '["postgres"]'
    ad.datadoghq.com/service.init_configs: '[{}]'
    ad.datadoghq.com/service.instances: |
      [
        {
          "dbm": true,
          "host": "<INSTANCE_ADDRESS>",
          "port": 5432,
          "username": "datadog",
          "password": "<UNIQUEPASSWORD>",
          "gcp": {
            "project_id": "<PROJECT_ID>",
            "instance_id": "<INSTANCE_ID>"
          }
        }
      ]      
spec:
  ports:
  - port: 5432
    protocol: TCP
    targetPort: 5432
    name: postgres

Validate

Run the Agent’s status subcommand and look for postgres under the Checks section. Or visit the Databases page to get started!

Example Agent Configurations

One agent connecting to multiple hosts

It is common to configure a single Agent host to connect to multiple remote database instances (see Agent installation architectures for DBM). To connect to multiple hosts, create an entry for each host in the Postgres integration config. In these cases, Datadog recommends limiting the number of instances per Agent to a maximum of 10 database instances to guarantee reliable performance.

init_config:
instances:
  - dbm: true
    host: example-service-primary.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    tags:
      - 'env:prod'
      - 'team:team-discovery'
      - 'service:example-service'
  - dbm: true
    host: example-service–replica-1.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    tags:
      - 'env:prod'
      - 'team:team-discovery'
      - 'service:example-service'
  - dbm: true
    host: example-service–replica-2.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    tags:
      - 'env:prod'
      - 'team:team-discovery'
      - 'service:example-service'
    [...]

Monitoring multiple databases on a database host

Use the database_autodiscovery option to permit the Agent to discover all databases on your host to monitor. You can specify include or exclude fields to narrow the scope of databases discovered. See the sample postgres.d/conf.yaml for more details.

init_config:
instances:
  - dbm: true
    host: example-service-primary.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    database_autodiscovery:
      enabled: true
      # Optionally, set the include field to specify
      # a set of databases you are interested in discovering
      include:
        - mydb.*
        - example.*
    tags:
      - 'env:prod'
      - 'team:team-discovery'
      - 'service:example-service'

Storing passwords securely

While it is possible to declare passwords directly in the Agent configuration files, it is a more secure practice to encrypt and store database credentials elsewhere using secret management software such as Vault. The Agent is able to read these credentials using the ENC[] syntax. Review the secrets management documentation for the required setup to store these credentials. The following example shows how to declare and use those credentials:

init_config:
instances:
  - dbm: true
    host: localhost
    port: 5432
    username: datadog
    password: 'ENC[datadog_user_database_password]'

Running custom queries

To collect custom metrics, use the custom_queries option. See the sample postgres.d/conf.yaml for more details.

init_config:
instances:
  - dbm: true
    host: localhost
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    custom_queries:
    - metric_prefix: employee
      query: SELECT age, salary, hours_worked, name FROM hr.employees;
      columns:
        - name: custom.employee_age
          type: gauge
        - name: custom.employee_salary
           type: gauge
        - name: custom.employee_hours
           type: count
        - name: name
           type: tag
      tags:
        - 'table:employees'

Monitoring relation metrics for multiple databases

In order to collect relation metrics (such as postgresql.seq_scans, postgresql.dead_rows, postgresql.index_rows_read, and postgresql.table_size), the Agent must be configured to connect to each database (by default, the Agent only connects to the postgres database).

Specify a single “DBM” instance to collect DBM telemetry from all databases. Use the database_autodiscovery option to avoid specifying each database name.

init_config:
instances:
  # This instance is the "DBM" instance. It will connect to the
  # all logical databases, and send DBM telemetry from all databases
  - dbm: true
    host: example-service-primary.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    database_autodiscovery:
      enabled: true
      exclude:
        - ^users$
        - ^inventory$
    relations:
      - relation_regex: .*
  # This instance only collects data from the `users` database
  # and collects relation metrics from tables prefixed by "2022_"
  - host: example-service-primary.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    dbname: users
    dbstrict: true
    relations:
      - relation_regex: 2022_.*
        relkind:
          - r
          - i
  # This instance only collects data from the `inventory` database
  # and collects relation metrics only from the specified tables
  - host: example-service-primary.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    dbname: inventory
    dbstrict: true
    relations:
      - relation_name: products
      - relation_name: external_seller_products

Collecting schemas

To enable this feature, use the collect_schemas option. You must also configure the Agent to connect to each logical database.

Use the database_autodiscovery option to avoid specifying each logical database. See the sample postgres.d/conf.yaml for more details.

init_config:
# This instance only collects data from the `users` database
# and collects relation metrics only from the specified tables
instances:
  - dbm: true
    host: example-service-primary.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    dbname: users
    dbstrict: true
    collect_schemas:
      enabled: true
    relations:
      - products
      - external_seller_products
  # This instance detects every logical database automatically
  # and collects relation metrics from every table
  - dbm: true
    host: example-service–replica-1.example-host.com
    port: 5432
    username: datadog
    password: '<PASSWORD>'
    database_autodiscovery:
      enabled: true
    collect_schemas:
      enabled: true
    relations:
      - relation_regex: .*

Working with hosts through a proxy

If the Agent must connect through a proxy such as the Cloud SQL Auth proxy, all telemetry is tagged with the hostname of the proxy rather than the database instance. Use the reported_hostname option to set a custom override of the hostname detected by the Agent.

init_config:
instances:
  - dbm: true
    host: localhost
    port: 5000
    username: datadog
    password: '<PASSWORD>'
    reported_hostname: example-service-primary
  - dbm: true
    host: localhost
    port: 5001
    username: datadog
    password: '<PASSWORD>'
    reported_hostname: example-service-replica-1

Install the Cloud SQL integration

To collect more comprehensive database metrics from Google Cloud, install the Cloud SQL integration (optional).

Troubleshooting

If you have installed and configured the integrations and Agent as described and it is not working as expected, see Troubleshooting

Further reading

Additional helpful documentation, links, and articles:

Basic Postgres IntegrationDOCUMENTATION