How to get JDBC traces with Arconia OpenTelemetry

[wimdeblauwe-volvo]

I am testing Arconia OpenTelemetry and I have it working inside my Spring Boot 3.5.7 appplication. I get Kafka and IBM MQ traces and manual traces from @Observed. However, I don't see JDBC traces.

I can get JDBC traces by adding this dependency:

<dependency>
      <groupId>net.ttddyy.observation</groupId>
      <artifactId>datasource-micrometer-spring-boot</artifactId>
      <version>1.2.0</version>
    </dependency>

Is this the intention to do it like this?

[ThomasVitale]

@wimdeblauwe-volvo thanks for giving Arconia OpenTelemetry a try. That is the current intention. At least, the Arconia BOM should manage the version for that dependency. I raised this feature request for doing that (and add documentation for it): #139

I'll also think about whether it could be included in the Arconia OpenTelemetry starter. It might not be a problem for Spring Boot 3, but with the new modularisation done in Spring Boot 4, that will not be a good idea at that point. Hopefully, spring-projects/spring-boot#39287 gets implemented and the instrumentation gets included in the Spring Boot project directly so that it's not necessary to handle this dependency separately and explicitly.

[wimdeblauwe-volvo]

I discovered an alternative that seems to give a bit nicer traces is adding this dependency:

      <dependency>
        <groupId>io.opentelemetry.instrumentation</groupId>
        <artifactId>opentelemetry-jdbc</artifactId>
      </dependency>

To make it work, you also need this BeanPostProcessor:

import io.opentelemetry.api.OpenTelemetry;
import io.opentelemetry.instrumentation.jdbc.datasource.JdbcTelemetry;
import javax.sql.DataSource;
import org.springframework.beans.BeansException;
import org.springframework.beans.factory.config.BeanPostProcessor;
import org.springframework.stereotype.Component;

@Component
public class OpenTelemetryJdbcTracingConfigurator implements BeanPostProcessor {

  private final OpenTelemetry openTelemetry;

  public OpenTelemetryJdbcTracingConfigurator(OpenTelemetry openTelemetry) {
    this.openTelemetry = openTelemetry;
  }

  @Override
  public Object postProcessAfterInitialization(Object bean, String beanName) throws BeansException {
    if( bean instanceof DataSource dataSource) {
      return JdbcTelemetry.create(openTelemetry).wrap(dataSource);
    }
    return BeanPostProcessor.super.postProcessAfterInitialization(bean, beanName);
  }
}

What I mean with "better" is that the trace is called "INSERT testdb.orders" for example instead of "dataSource". The span attributes look like this:

| db.connection_string | "oracle:thin://localhost:1521" |
| db.name              | "testdb" |
| db.operation         | "INSERT" |
| db.sql.table         | "orders" |
| db.statement         | "insert into orders (car_type,quantity,id) values (?,?,?)" |
| db.system            | "oracle" |
| server.address       | "localhost" |
| server.port          | 1521 |

With datasource-micrometer-spring-boot, the trace is called "dataSource: query" with these attributes:

| jdbc.datasource.driver | "oracle.jdbc.OracleDriver" |
| jdbc.datasource.name   | "dataSource" |
| jdbc.datasource.pool   | "HikariPool-1" |
| jdbc.query[0]          | "insert into orders (car_type,quantity,id) values (?,?,?)" |
| jdbc.row-affected      | "1" |
| peer.service           | "dataSource" |