jaeger-client-python

You can force a trace to be sampled using two methods:

1. Programmatically: Set the sampling.priority tag on a span.
2. Via HTTP Headers: Include the jaeger-debug-id header in an incoming request. This ensures the trace is sampled in 'debug' mode and the root span is tagged with the provided ID, allowing you to find it in the Jaeger UI.

# Programmatic approach
from opentracing.ext import tags as ext_tags
span.set_tag(ext_tags.SAMPLING_PRIORITY, 1)

# HTTP Header approach
curl -H "jaeger-debug-id: some-correlation-id" http://myhost.com

The recommended pattern for production is to wrap initialization in a function. initialize_tracer() sets the global opentracing.tracer. If you need to create additional tracers for remote services, use the new_tracer() method.

from jaeger_client import Config

def init_jaeger_tracer(service_name='your-app-name'):
    config = Config(config={}, service_name=service_name, validate=True)
    return config.initialize_tracer()

Do not initialize the tracer during module import, as this may cause a deadlock. Instead, define a function that returns a tracer and call it explicitly after all imports are completed.

If using gevent.monkey in asyncio-based applications (Python 3+), you may need to pass the current event loop explicitly to initialize_tracer.

from tornado import ioloop
from jaeger_client import Config

config = Config(config={}, service_name='your-app-name', validate=True)
config.initialize_tracer(io_loop=ioloop.IOLoop.current())

Install the Jaeger client library using pip.

pip install jaeger-client

To support Zipkin B3 propagation (using X-B3-* headers), set the propagation configuration property to 'b3'. Ensure your framework (like Flask or Django) provides these headers in lowercase.

# Example configuration key
config = {
    'propagation': 'b3'
}

If running Jaeger in a separate container (e.g., using the all-in-one Docker image), you can specify the remote agent's location using the local_agent configuration key. Note that Jaeger sends spans over UDP, which does not guarantee delivery.

    config = Config(
        config={
            'sampler': {
                'type': 'const',
                'param': 1,
            },
            'local_agent': {
                'reporting_host': 'your-reporting-host',
                'reporting_port': 'your-reporting-port',
            },
            'logging': True,
        },
        service_name='your-app-name',
        validate=True,
    })

You can integrate Prometheus metrics by using the PrometheusMetricsFactory. This allows you to monitor internal Jaeger client metrics. Use the service_name_label argument to tag metrics with a specific service name, making it easier to distinguish metrics from different services.

from jaeger_client.metrics.prometheus import PrometheusMetricsFactory

config = Config(
        config={},
        service_name='your-app-name',
        validate=True,
        metrics_factory=PrometheusMetricsFactory(service_name_label='your-app-name')
)
tracer = config.initialize_tracer()

The jaeger service (using the jaegertracing/all-in-one image) can be configured via environment variables:

• COLLECTOR_ZIPKIN_HTTP_PORT: Sets the port for the Zipkin collector compatibility (e.g., 9411).
• LOG_LEVEL: Sets the logging verbosity (e.g., debug).

jeager:
  image: jaegertracing/all-in-one
  environment:
    - COLLECTOR_ZIPKIN_HTTP_PORT=9411
    - LOG_LEVEL=debug