Pipelines

pgagroal supports 3 different pipelines that determine how connections are managed and what features are available.

The pipeline is defined in pgagroal.conf under the setting:

pipeline = auto

pgagroal will choose either the performance or the session pipeline based on the configuration settings by default.

Performance Pipeline

The performance pipeline is the fastest pipeline as it is a minimal implementation of the pipeline architecture.

However, it doesn't support Transport Layer Security (TLS), failover support and the disconnect_client setting.

A DISCARD ALL query is run after each client session.

Configuration

Select the performance pipeline by:

pipeline = performance

Use Cases

The performance pipeline is ideal for:

• High-throughput applications
• Scenarios where maximum performance is critical
• Environments where TLS is not required
• Simple connection pooling needs

Limitations

• No Transport Layer Security (TLS) support
• No failover support
• No disconnect_client setting support
• Minimal feature set

Session Pipeline

The session pipeline supports all features of pgagroal.

A DISCARD ALL query is run after each client session.

Configuration

Select the session pipeline by:

pipeline = session

Features

The session pipeline supports:

• Transport Layer Security (TLS)
• Failover functionality
• All configuration options
• Complete feature set

Use Cases

The session pipeline is ideal for:

• Production environments requiring full features
• Applications needing TLS encryption
• Environments with failover requirements
• Complex connection pooling scenarios

Transaction Pipeline

The transaction pipeline will release the connection back to the pool after each transaction completes. This feature will support many more clients than there are database connections.

Configuration

Select the transaction pipeline by:

pipeline = transaction

Features

• Connection released after each transaction
• Supports many more clients than database connections
• Automatic transaction boundary detection
• Rollback handling for failed transactions

Use Cases

The transaction pipeline is ideal for:

• Applications with many short transactions
• Microservices architectures
• High-concurrency scenarios with brief database interactions
• Applications that can handle connection state loss between transactions

Considerations

• Application must handle loss of connection state between transactions
• Prepared statements are not preserved across transactions
• Temporary tables and other session-specific objects are not available
• May require application code changes

Pipeline Comparison

Feature	Performance	Session	Transaction
Speed	Fastest	Fast	Moderate
TLS Support	No	Yes	Yes
Failover Support	No	Yes	Yes
Connection Reuse	Session-based	Session-based	Transaction-based
Client Capacity	Limited by pool size	Limited by pool size	High
State Preservation	Session	Session	None
Complexity	Low	Medium	High

Choosing the Right Pipeline

Performance Pipeline

Choose when:

• Maximum performance is required
• TLS is not needed
• Simple connection pooling is sufficient
• Failover is handled externally

Session Pipeline

Choose when:

• Full feature set is required
• TLS encryption is needed
• Failover support is required
• Standard connection pooling behavior is desired

Transaction Pipeline

Choose when:

• High client concurrency is needed
• Connections are used for short transactions
• Application can handle stateless connections
• Database connection limits are a constraint

Configuration Examples

High-Performance Setup

[pgagroal]
pipeline = performance
max_connections = 100
validation = off

Production Setup with TLS

[pgagroal]
pipeline = session
max_connections = 50
tls = on
tls_cert_file = /path/to/cert.pem
tls_key_file = /path/to/key.pem
failover = on
failover_script = /path/to/failover.sh

High-Concurrency Setup

[pgagroal]
pipeline = transaction
max_connections = 20
# Support many more clients than connections