Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

Warning

The Apache HTrace project has retired on 2018-04-11 and is not developed anymore. The Apache CXF retired the Apache HTrace integration, starting from 3.3.0 release. 

Overview

Apache HTrace is a tracing framework intended for use with distributed systems written in java. Since version 3.1.3, Apache CXF fully supports integration with Apache HTrace, both on client side and server side. This section gives a complete overview on how distributed tracing support is supported in JAX-RS applications built on top of Apache CXF.

...

Distributed tracing is additional instrumentation layer on top of new or existing applications. In terms of distributed tracing, span represents a basic unit of work. For example, executing database query is a span. Spans are identified by a unique 128-bit ID. Spans also have other data, such as descriptions, timelines, key-value annotations, the ID of the span that caused them (parent), and process/tracer ID’s (normally IP address and process name). Spans are started and stopped, and they keep track of their timing information. Once span is created, it should be stopped at some point in the future. In turn, trace is a set of spans forming a tree-like structure. For example, if you are running a JAX-RS service, a trace might be formed by a PUT request and downstream work.

From implementation prospectiveperspective, and in context of Java applications, spans are attached to their threads (in general, thread which created the span should close it). However it is possible to transfer spans from thread to thread in order to model a complex execution flows. It is also possible to have many spans in the same thread, as long as they are properly created and closed. In the next sections we are going to see the examples of that.

Another two important concepts of in context of distributed tracing are span receivers and samplers. Essentially, all spans (including start/stop time, key/value annotations, timelines, ..) should be persisted (or collected) somewhere. Span receiver is a collector within a process that is the destination of spans when a trace is running (it could be a console, local file, data store, ...). Apache HTrace provides span receivers for Apache HBase, Apache Flume and Twitter Zipkin. From other side, samplers allow to control the frequency of the tracing (all the time, never, probability driven, ...). Using the sampler is the way to minimize tracing overhead (or just amount of traces) by limiting them to particular conditions.

Distributed Tracing in Apache CXF using Apache HTrace

Apache CXF is a very popular framework for building services and web APIs. No doubts, it is going to play even more important role in context of microservices architecture letting developers to quickly build and deploy individual JAX-RS/JAX-WS services. As it was just mentioned before, distributed tracing is an essential technique to monitor the application as whole, breaking the request to individual service traces as it goes through and crosses the boundaries of threads, processes and machines.

The current integration of distributed tracing in Apache CXF supports Apache HTrace (4.x+ release branch) only in JAX-RS 2.x applications. From high-level prospective, it consists of three main parts:

...

By default, HTraceProvider will try to pass the currently active span through HTTP headers on each service invocation. If there is no active spanspans, the new span will be created and passed through HTTP headers on per-invocation basis. Essentially, just registering the HTraceProvider HTraceClientProvider on the client and HTraceClientProvider and HTraceProvider on the server is enough to have tracing context to be properly passed everywhere. The only configuration part which is necessary are span receiver(s) and sampler(s).

It is also worth to mention the way Apache CXF attaches the description to spans. With regards to the client integration, the description becomes a full URL being invoked prefixed by HTTP method, for example: GET http://localhost:8282/books. On the server side integration, the description becomes a relation relative JAX-RS resource path prefixed by HTTP method, f.e.: GET books, POST book/123

Configuring Client
Anchor
configure.client
configure.client

There are a couple of way ways the JAX-RS client could be configured, depending on the client implementation. Apache CXF provides its own WebClient which could be configured just like that (in future versions, there would be a simpler ways to do that using client specific features):

Code Block
java
java
final Map<String, String> properties = new HashMap<String, String>();
properties.put(Tracer.SPAN_RECEIVER_CLASSES_KEY, ...);
properties.put(Tracer.SAMPLER_CLASSES_KEY, ...);

/**
 * For example:
 *
 * properties.put(Tracer.SPAN_RECEIVER_CLASSES_KEY, StandardOutSpanReceiver.class.getName());
 * properties.put(Tracer.SAMPLER_CLASSES_KEY, AlwaysSampler.class.getName());
 */
        
final Tracer tracer = new Tracer.Builder("web-client")
    .name("webclient")
    .conf(HTraceConfiguration.conf(HTraceConfiguration.fromMap(properties))
    .build();
                
final HTraceClientProvider provider = new HTraceClientProvider(tracer);
final Response response = WebClient
    .create("http://localhost:9000/catalog", Arrays.asList(provider))
    .accept(MediaType.APPLICATION_JSON)
    .get();

...

Code Block
java
java
final Map<String, String> properties = new HashMap<String, String>();
properties.put(Tracer.SPAN_RECEIVER_CLASSES_KEY, ...);
properties.put(Tracer.SAMPLER_CLASSES_KEY, ...);

/**
 * For example:
 *
 * properties.put(Tracer.SPAN_RECEIVER_CLASSES_KEY, StandardOutSpanReceiver.class.getName());
 * properties.put(Tracer.SAMPLER_CLASSES_KEY, AlwaysSampler.class.getName());
 */
        
final Tracer tracer = new Tracer.Builder()
    .name("jaxrs-client")
    .conf(HTraceConfiguration.fromMap(properties))
    .build();
                
final HTraceClientProvider provider = new HTraceClientProvider(tracer);
final Client client = ClientBuilder.newClient().register(provider);

final Response response = client
    .target("http://localhost:8282/rest/api/people9000/catalog")
    .request()
    .accept(MediaType.APPLICATION_JSON)
    .get();

...

Server configuration is a bit simpler than the client one thanks to the feature class available, HTraceFeature. Depending on the way the Apache CXF is used to configure JAX-RS services, it could be part of JAX-RS application configuration, for example:

...

Code Block
java
java
final Map<String, String> properties = new HashMap<String, String>();
properties.put(Tracer.SPAN_RECEIVER_CLASSES_KEY, ...);
properties.put(Tracer.SAMPLER_CLASSES_KEY, ...);

final JAXRSServerFactoryBean factory = 
    RuntimeDelegate.getInstance().createEndpoint(/* application instance */, JAXRSServerFactoryBean.class);
factory.setFeatures(Arrays.< Feature >asList(new HTraceFeature(HTraceConfiguration.fromMap(properties), "tracer")));
...
return factory.create();

...

In the following subsections we are going to walk through many different scenarios to illustrate the distributed tracing in action, starting from the simplest ones and finishing with asynchronous JAX-RS services. All examples assume that configuration has been done (see please Configuring Client and Configuring Server sections above).

Example #1: Client and Server with default distributed tracing configured

...

In this respect, there is no difference from the caller prospective however a bit more work is going under the hood to transfer the active tracing span from JAX-RS client request filter to client response filter as in general those are executed in different threads (similarly to server-side asynchronous JAX-RS resource invocation). The actual invocation of the request by the client (with process name jaxrsclient/192.168.0.100) and consequent invocation of the service on the server side (process name jaxrsserver/192.168.0.100) is going to generate the following sample traces:

Distributed Tracing Apache HTrace and JAX-WS support

Distributed tracing in the Apache CXF is build primarily around JAX-RS 2.x implementation. However, JAX-WS is also supported but it requires to write some boiler-plate code and use Apache HTrace API directly (the JAX-WS integration is going to be enhanced in the nearest future). Essentially, from the server-side prospective the in/out interceptors, HTraceStartInterceptor and HTraceStopInterceptor respectively, should be configured as part of interceptor chains. The span receiver should be configured manually though, using Apache HTrace API, for example:

...

Info

Configuring right phases for interceptors is very important. The recommended phase for in-interceptor is PRE_INVOKE while for out-interceptor is PREPOST_MARSHAL. If wrong phases are being used, response or/and request headers, status code, etc could be ignored or not processed.

...

Code Block
java
java
JaxWsProxyFactoryBean factory = new JaxWsProxyFactoryBean();
...
factory.getOutInterceptors().add(new HTraceClientStartInterceptor(sampler));
factory.getInInterceptors().add(new HTraceClientStopInterceptor());
...
factory.create();

Propagating Trace Details To Logs

In order to have better correlation between ongoing traces and logs, Apache CXF since 3.1.11 / 3.2.0 releases distributes a helpful extension for Logback users, LogbackSpanConverter. This converter can be used to complement log records with current trace details, such as tracer id and span id. For example, here is a simple logback.xml configuration file.

Code Block
xml
xml
<?xml version="1.0" encoding="UTF-8"?>
<configuration scan="true" scanPeriod="5 seconds">
    <conversionRule conversionWord="trace" converterClass="org.apache.cxf.tracing.htrace.ext.LogbackSpanConverter" />
    <appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender">
        <encoder>
            <pattern>[%level] [%trace] %d{yyyy-MM-dd HH:mm:ss.SSS} %logger{36} %msg%n</pattern>
        </encoder>
    </appender>
    <root level="INFO">
        <appender-ref ref="STDOUT" />
    </root>
</configuration>

In this case the tracing details will be propagated to each log record in following format: <tracer_id>, span: <span id>. For example:

Code Block
bash
bash
[INFO] [-, -] 2017-03-11 14:40:13.603 org.eclipse.jetty.server.Server Started @2731ms
[INFO] [tracer-server/192.168.0.101, span: 6d3e0d975d4c883cce12aee1fd8f3e7e] 
2017-03-11 14:40:24.013 com.example.rs.PeopleRestService Getting all employees
[INFO] [tracer-server/192.168.0.101, span: 6d3e0d975d4c883c7592f4c2317dec22] 
2017-03-11 14:40:28.017 com.example.rs.PeopleRestService Looking up manager in the DB database

The special [-, -] placeholder indicates that no trace details are being available at the moment of logging the record.

Accessing Apache HTrace APIs

The Apache CXF  abstracts as much of the tracer-specific APIs behind TracerContext as possible. However, sometimes there is a need to get access to Apache HTrace APIs in order to leverages the available instrumentations. To make it possible, TracerContext has a dedicated unwrap method which returns underlying Tracer instance.

Code Block
java
java
@GET
@Path("/search")
@Produces(MediaType.APPLICATION_JSON)
public JsonObject search(@QueryParam("q") final String query, @Context final TracerContext tracing) throws Exception {
    final Tracer tracer = tracing.unwrap(Tracer.class);
    // ...
}

Future Work

The Apache CXF is very proud to offer Apache HTrace integration. At the current stage, it was a conscious decision to keep the minimal API and provide the set of necessary features only. However, there is a strong commitment to evolve not only Apache HTrace integration, but the distributed tracing support in general.