Connect PostgreSQL databases to PMM¶

Connect your PostgreSQL databases—whether self-hosted or running in the cloud—to Percona Monitoring and Management (PMM) for comprehensive performance monitoring and analysis.

PMM Client supports collecting metrics from PostgreSQL-based database systems:

For monitoring Amazon RDS PostgreSQL instances, see Connect Amazon RDS instance.

Setup process at a glance

These are the high-level steps for configuring PostgreSQL monitoring in PMM:

Prerequisites: Ensure PMM Server is running and PMM Client is installed
Create PMM user: CREATE USER pmm WITH SUPERUSER ENCRYPTED PASSWORD 'StrongPassword'
Configure extension: Set up pg_stat_statements or pg_stat_monitor
Add service: Use PMM UI or command line to add the PostgreSQL instance
Verify connection: Check PMM Inventory and dashboards for data.

Before you start¶

Check that:

PMM Server is installed and running with a known IP address accessible from the client node.
PMM Client is installed and the nodes are registered with PMM Server.
You have superuser (root) access on the client host.
You have superuser access to any database servers that you want to monitor.

PMM follows PostgreSQL’s end-of-life policy. For specific details on supported platforms and versions, see Percona’s Software Platform Lifecycle page.

Create a database account for PMM¶

PMM requires a dedicated database account with appropriate permissions to collect metrics effectively. We recommend creating a PMM database account that can connect to the postgres database with the SUPERUSER role.

Standard PostgreSQLAmazon RDS/Aurora PostgreSQL

To create a user with the SUPERUSER role:

Create a user with SUPERUSER role:

CREATE USER pmm WITH SUPERUSER ENCRYPTED PASSWORD 'strong_password';

Edit the pg_hba.conf file to enable local login:

local   all             pmm                                md5
# TYPE  DATABASE        USER        ADDRESS                METHOD

For RDS instances where SUPERUSER cannot be assigned directly:

Create the user:

CREATE USER pmm WITH ENCRYPTED PASSWORD 'strong_password';

Grant the rds_superuser role:
```
GRANT rds_superuser TO pmm;
```
Optionally, set a connection limit (only if the user is not a SUPERUSER):
```
ALTER USER pmm CONNECTION LIMIT 10;
```

Configure authentication¶

After creating the database user, complete the configuration:

Reload the PostgreSQL configuration:

su - postgres
psql -c "select pg_reload_conf()"

Verify the PMM user can connect locally:
```
psql postgres pmm -c "\conninfo"
```
Enter the password for the pmm user when prompted:

Choose a monitoring extension¶

Decide which PostgreSQL monitoring extensions to use, and configure your database server for it.

Choose:

pg_stat_monitor when you need comprehensive monitoring capabilities with more detailed insights into query performance:
pg_stat_statements when you need a lightweight, built-in solution with minimal overhead

Aspect	pg_stat_statements	pg_stat_monitor
Origin & packaging	• Official PostgreSQL extension • Part of `postgresql-contrib` package on Linux	• Developed by Percona • Requires separate installation
Key features	✔ Basic query statistics ✔ Simple aggregation ✔ Minimal overhead ✔ Query timing and execution counts	✔ Enhanced metrics collection ✔ Bucket-based aggregation ✔ Query examples ✔ Histogram data ✔ Includes all pg_stat_statements features
Best for	✔ Development environments ✔ Simple monitoring needs ✔ Resource-constrained servers	✔ Production environments ✔ Detailed query analysis ✔ Performance tuning
Installation complexity	⚠ Low	⚠ Medium
Benefits	• Part of official PostgreSQL • Minimal overhead • Simple to set up and use	• Builds on pg_stat_statements features • Bucket-based time-series analysis • Query examples for troubleshooting • More accurate performance data
Drawbacks	• No aggregated statistics or histograms • No Query Examples • Limited metrics collection	• Slightly higher resource overhead • Requires separate installation • More complex configuration

For a more detailed comparison of extensions, see the pg_stat_monitor documentation.

Configure monitoring extension¶

pg_stat_monitorpg_stat_statements

pg_stat_monitor is Percona’s advanced PostgreSQL monitoring extension that enhances observability with detailed query metrics and improved aggregation. It is compatible with PostgreSQL and Percona Distribution for PostgreSQL versions 11 through 15.

Install the extension:
- For Percona Distribution for PostgreSQL: Install via your Linux package manager after setting up the Percona repository, making sure to use the command with your actual PostgreSQL version: apt install -y pg-stat-monitor-15
- For Standard PostgreSQL: Download and compile from source code.

Configure PostgreSQL settings in postgresql.conf:

# Add to shared libraries
shared_preload_libraries = 'pg_stat_monitor'

# Required for PMM
pg_stat_monitor.pgsm_query_max_len = 2048

# Recommended settings
pg_stat_monitor.pgsm_normalized_query = 1

Using both extensions?

If using with pg_stat_statements, list it first:

shared_preload_libraries = 'pg_stat_statements, pg_stat_monitor'

Restart your PostgreSQL instance:
```
systemctl restart postgresql
```

Create the extension:

psql -d postgres -c "CREATE EXTENSION pg_stat_monitor;"

Verify the installation:
```
SELECT pg_stat_monitor_version();
```

About bucket-based aggregation¶

pg_stat_monitor uses buckets to collect and aggregate statistics:

Each bucket collects data for a configurable time period
When a bucket expires, data moves to the next bucket in the chain
When all buckets are full, the oldest bucket is reused
If a bucket fills before expiring, excess data is discarded

pg_stat_statements is the built-in PostgreSQL extension for tracking query performance, available as part of the postgresql-contrib package.

Install the required package:
- Debian/Ubuntu: apt install -y postgresql-contrib
- Red Hat/CentOS: yum install -y postgresql-contrib

Add these lines to your postgresql.conf file:

shared_preload_libraries = 'pg_stat_statements'
track_activity_query_size = 2048 # Increase tracked query string size
pg_stat_statements.track = all   # Track all statements including nested
track_io_timing = on             # Capture read/write stats

Restart the PostgreSQL server:
```
systemctl restart postgresql
```
Create the extension:
```
psql postgres postgres -c "CREATE EXTENSION pg_stat_statements SCHEMA public"
```
Best practice

Create the extension in the postgres database to access statistics from all databases without configuring each one individually.

Add service to PMM¶

After configuring your database server with the appropriate extension, you need to add it as a service to PMM. You can do this either through the PMM user interface or via the command line.

Via web UIVia command line

To add the service from the user interface:

Go to PMM Configuration > Add Service > PostgreSQL.
Enter or select values for the fields.
Click Add service.
If using TLS, check Use TLS for database connections and fill in your TLS certificates and key.
For TLS connection, make sure SSL is configured in your PostgreSQL instance.

Make sure SSL is enabled in the server configuration file postgresql.conf, and that hosts are allowed to connect in the client authentication configuration file pg_hba.conf. See PostgreSQL documentation on Secure TCP/IP Connections with SSL.

Basic setupCustom service nameUNIX socket connectionTLS connection

Add an instance with default node name:

pmm-admin add postgresql \
--username=pmm \
--password=password \
--server-url=https://admin:admin@X.X.X.X:443 \
--server-insecure-tls

The service name will be automatically generated based on the node name.

Add an instance with a specified service name:

pmm-admin add postgresql \
--username=pmm \
--password=password \
--server-url=https://admin:admin@X.X.X.X:443 \
--server-insecure-tls \
--service-name=SERVICE-NAME

Add an instance using a UNIX socket:

pmm-admin add postgresql --socket=/var/run/postgresql

Where:

/var/run/postgresql: Directory containing the socket

Add an instance with TLS security:

pmm-admin add postgresql --tls \
--tls-cert-file=PATHTOCERT \
--tls-ca-file=PATHTOCACERT \
--tls-key-file=PATHTOKEY \
--host=HOST \
--port=PORT \
--username=USER \
--service-name=SERVICE-NAME

where:

PATHTOCERT: Path to client certificate file
PATHTOCACERT: Path to certificate authority file
PATHTOKEY: Path to client key file
HOST: Instance hostname or IP
PORT: PostgreSQL service port number
USER: Database user allowed to connect via TLS (should match the CN in the client certificate)
SERVICE-NAME: Name to give to the service within PMM

Configure auto-discovery¶

Auto-discovery dynamically identifies all databases in your PostgreSQL instance. This feature helps balance comprehensive monitoring with resource efficiency.

Performance impact

Limiting auto-discovery may result in fewer metrics being captured from the non-primary databases. Ensure that you set the limit appropriately:

High limits may impact performance by creating too many connections
Low limits may result in missing metrics from non-primary databases

Via command lineVia web UI

The pmm-admin flag controls Auto-discovery behavior:

--auto-discovery-limit=XXX

How the limit works:

If number of databases > Auto-discovery limit: Auto-discovery is OFF
If number of databases <= Auto-discovery limit: Auto-discovery is ON
If Auto-discovery limit is not defined: Default value is 0 (server-defined with limit 10)
If Auto-discovery limit < 0: Auto-discovery is OFF

Example:

pmm-admin add postgresql \
--username="pmm" \
--password="password" \
--auto-discovery-limit=10

If your PostgreSQL instance has 11 databases, automatic discovery will be disabled.

By default, Auto-discovery is enabled with a server-defined limit of 10 databases.

Auto-discovery Enabled

When you select Disabled, the Auto-discovery limit will be set to -1.

Auto-discovery Disabled

For a custom value, select Custom and enter your preferred limit.

Auto-discovery Custom

Check the service¶

After adding a PostgreSQL service, verify that it’s properly connected and sending data to PMM.

Via command lineVia web UICheck the dashboard data

Run this command to view all services:

pmm-admin inventory list services

Docker environments

If using Docker, use:

docker exec pmm-client pmm-admin inventory list services

Look for your PostgreSQL service in the output and verify that its status is “RUNNING”.

Use the UI to confirm that your service was added and is actively monitored:

Select Configuration > Inventory.
In the Services tab, verify that Service name matches what you configured, Address points to your PostgreSQL instance and Status shows as “Active”.
In the Options column, expand the Details section to check that agents are properly registered and that the expected data source is being used.

Ensure PostgreSQL metrics are flowing and visualized correctly:

Open the PostgreSQL Instance Summary dashboard.
Select your service name from the dropdown.
Verify that metrics are being displayed.
Check that the graphs are updating with current data.

Running custom queries¶

The PostgreSQL exporter can execute custom queries to collect additional metrics beyond what PMM provides by default.

Configuration basicsQuery definition formatMetric types

Custom queries must be defined in this directory on the host where the exporter is running:

/usr/local/percona/pmm/collectors/custom-queries/postgresql

Three resolution directories are available:

Directory	Execution Frequency
`high-resolution/`	Every 5 seconds
`medium-resolution/`	Every 10 seconds
`low-resolution/`	Every 60 seconds

Choose the appropriate directory based on how frequently you need the data.

Create YAML files in the appropriate resolution directory with this structure:

query_name:
  query: "SELECT statement goes here"
  master: true|false  # Whether to run only on the master
  metrics:
    - metric_name:
        usage: "GAUGE|COUNTER|LABEL|MAPPEDMETRIC|DURATION"
        description: "Human-readable description of the metric"