OpenTelemetry

The foundations-jdbc-otel module wires OpenTelemetry into the existing QueryListener mechanism, producing spans and metrics for every query. When no OTel SDK is configured, the listener does nothing.

Setup

Add the foundations-jdbc-otel module to your project alongside the core library. Then create an OtelQueryListener and attach it to your transactor:

Kotlin

val config =
    PgConfig.builder("localhost", 5432, "mydb", "user", "pass").build()

val telemetryConfig = TelemetryConfig.builder(config).build()

val otelListener: QueryListener =
    OtelQueryListener.create(GlobalOpenTelemetry.get(), telemetryConfig)

val tx = Transactor.create(config).withListener(otelListener)

Java

PgConfig config = PgConfig.builder("localhost", 5432, "mydb", "user", "pass").build();

TelemetryConfig telemetryConfig = TelemetryConfig.builder(config).build();

QueryListener otelListener = OtelQueryListener.create(GlobalOpenTelemetry.get(), telemetryConfig);

Transactor tx = Transactor.create(config).withListener(otelListener);

Scala

val config =
  PgConfig
    .builder("localhost", 5432, "mydb", "user", "pass")
    .build()

val telemetryConfig = TelemetryConfig.builder(config).build()

val otelListener: QueryListener =
  OtelQueryListener.create(GlobalOpenTelemetry.get(), telemetryConfig)

val tx = Transactor.create(config).withListener(otelListener)

Show entire file

TelemetryConfig.builder(config) extracts the database system, host, port, and database name from the DatabaseConfig. You can also build it manually:

var telemetryConfig = TelemetryConfig.builder(DatabaseSystem.POSTGRESQL)
    .serverAddress("db.example.com")
    .serverPort(5432)
    .dbNamespace("orders")
    .build();

Span attributes

Each query produces a span with attributes following OTel Semantic Conventions v1.39.0:

Attribute	Source
db.system.name	From TelemetryConfig
db.query.text	Raw SQL text (when recordQueryText is enabled)
server.address / server.port	From config
db.namespace	Database name from config
error.type	Exception class on failure

The span name is the operation's .named() value if set, otherwise "DB query".

TelemetryConfig options

Option	Default	Description
recordQueryText	true	Include db.query.text attribute

No ThreadLocal

The listener uses backdated spans rather than ThreadLocal state. beforeQuery is a no-op. afterQuery and failedQuery each create a complete span by computing Instant.now() - elapsed. This is safe for virtual threads, nested queries, and composed listeners.

Pool metrics

If you use HikariCP, register connection pool metrics:

Kotlin

val poolMetrics: PoolMetrics =
    PoolMetrics.register(GlobalOpenTelemetry.get(), dataSource, Optional.of("main-pool"))

fun onShutdown() {
    poolMetrics.close()
}

Java

PoolMetrics poolMetrics =
    PoolMetrics.register(GlobalOpenTelemetry.get(), dataSource, Optional.of("main-pool"));

void onShutdown() {
  poolMetrics.close();
}

Scala

val poolMetrics: PoolMetrics =
  PoolMetrics.register(GlobalOpenTelemetry.get(), dataSource, Optional.of("main-pool"))

def onShutdown(): Unit =
  poolMetrics.close()

Show entire file

Metrics follow the OTel database pool semantic conventions:

Metric	Description
db.client.connection.count	Active and idle connections (by state)
db.client.connection.pending_requests	Threads waiting for a connection
db.client.connection.max	Maximum pool size

Grafana dashboard

A pre-built Grafana dashboard is included at foundations-jdbc-otel/grafana/foundations-jdbc-dashboard.json. Import it into Grafana and configure:

• Tempo / Jaeger datasource for trace panels (latency percentiles, slow queries)
• Prometheus datasource for metric panels (throughput, pool status)

Panels include:

• Query latency p50/p95/p99 by operation
• Error rate
• Throughput by database system
• Connection pool active/idle/pending
• Pool usage percentage gauge
• Slow queries table