This is a legacy repository with com.uber.jaeger
package name. It is maintained for bug fixes.
Issues and pull requests with new functionality please submit to jaeger-client-java.
Please see CONTRIBUTING.md.
Click through for more detailed docs on specific modules.
All artifacts are published to Maven Central. Snapshot artifacts are also published to Sonatype. Follow these instructions to add the snapshot repository to your build system.
Add the required dependencies to your project. Usually, this would only be the add-ons you require. Please use the latest version:
For e.g, to depend on the core jaeger library, you'd include the following
<dependency>
<groupId>com.uber.jaeger</groupId>
<artifactId>jaeger-core</artifactId>
<version>$jaegerVersion</version>
</dependency>
Jaeger client uses org.apache.thrift:libthrift:0.9.2
. If your project depends on a different
version of libthrift
, it is recommended that you use the shaded jaeger-thrift
jar we publish
which packages it's own libthrift
.
To depend on the shaded jar, add the following to your maven build. Note that this is only supported for a jaeger version >= 0.15.0
<dependencyManagement>
<dependencies>
<dependency>
<groupId>com.uber.jaeger</groupId>
<artifactId>jaeger-thrift</artifactId>
<classifier>thrift92</classifier>
<version>$jaegerVersion</version>
</dependency>
</dependencies>
</dependencyManagement>
Please see jaeger-core/README.
jaeger-context
defines
ThreadLocalTraceContext
implementation of TraceContext
that can be used for propagating the current tracing Span
throughout the application without changing the application's code. However, if the application
is starting new threads or is using thread pools, the thread-local context is not going to be
carried over into the execution in the next thread. To maintain context propagation,
a wrapper TracedExecutorService
is provided that automatically transfers the context
onto the new threads.
ExecutorService instrumentedExecutorService = TracingUtils.tracedExecutor(wrappedExecutorService);
When testing tracing instrumentation it is often useful to make sure that all spans are being captured, which is not the case in production configurations where heavy sampling is applied by default. The following configuration can be provided to affect which sampling is applied to the new traces:
sampler:
type: const # can either be const, probabilistic, or ratelimiting
param: 1 # can either be an integer, a double, or an integer
The valid values for type
are:
const
: configures a sampler that always makes the same decision
for new traces depending on the param
: always no for param=0
,
always yes otherwise.probabilistic
: configures a sampler that samples traces with
probability equal to param
(must be between 0.0
and 1.0
)ratelimiting
: configures a samlper that samples traces with a
certain rate per second equal to param
The OpenTracing API defines a sampling.priority
standard tag that
can be used to affect the sampling of a span and its children:
import io.opentracing.tag.Tags;
Tags.SAMPLING_PRIORITY.set(span, 1);
Jaeger Tracer also understands a special HTTP Header jaeger-debug-id
,
which can be set in the incoming request, e.g.
curl -H "jaeger-debug-id: some-correlation-id" http://myhost.com
When Jaeger sees this header in the request that otherwise has no tracing context, it ensures that the new trace started for this request will be sampled in the "debug" mode (meaning it should survive all downsampling that might happen in the collection pipeline), and the root span will have a tag as if this statement was executed:
span.setTag("jaeger-debug-id", "some-correlation-id")
This allows using Jaeger UI to find the trace by this tag.