camel-zipkin

Available as of Camel 2.18

The camel-zipkin component is used for tracing and timing incoming and outgoing Camel messages using zipkin. Events (span) are captured for incoming and outgoing messages being sent to/from Camel. This means you need to configure which which Camel endpoints that maps to Zipkin service names.

The mapping can be configured using:

• route ID - A Camel route id
• endpoint URL - A Camel endpoint URL
  

For both kinds you can use wildcards and regular expressions to match, which is using the rules from Intercept. To match all Camel messages you can use * in the pattern and configure that to the same service name. If no mapping has been configured then Camel will fallback and use endpoint URIs as service names. However its recommended to configure service mappings so you can use human logic names instead of Camel endpoint URIs in the names.

Camel will auto-configure a ScribeSpanCollector if no SpanCollector explicit has been configured, and if the hostname and port to the span collector has been configured as environment variables:

• ZIPKIN_COLLECTOR_THRIFT_SERVICE_HOST - The hostname
• ZIPKIN_COLLECTOR_THRIFT_SERVICE_PORT - The port number

This makes it easy to use camel-zipkin in container platforms where the platform can run your application in a Linux container where service configurations are provided as environment variables.

Options

You can configure the following options on ZipkinEventNotifier:

Example

To enable camel-zipkin you need to configure first

ZipkinTracer zipkin = new ZipkinTracer();

// Configure the scribe span collector with the hostname and port for the Zipkin Collector Server 
zipkin.setSpanCollector(new ScribeSpanCollector("192.168.90.100", 9410);

// ...then add zipkin to the CamelContext
zipkin.init(camelContext);

The configuration about will the trace all incoming and outgoing messages in Camel routes. 

To use ZipkinTracer in XML all you need to do is to setup scribe an the Zipkin tracer as <bean> and then they are automatic discovered and used by Camel.

  <!-- configure the scribe span collector with the hostname and port for the Zipkin Collector Server -->
  <bean id="scribe" class="com.github.kristofa.brave.scribe.ScribeSpanCollector">
    <constructor-arg index="0" value="192.168.90.100"/>
    <constructor-arg index="1" value="9410"/>
  </bean>

  <!-- setup zipkin tracer -->
  <bean id="zipkinTracer" class="org.apache.camel.zipkin.ZipkinTracer">
    <property name="serviceName" value="dude"/>
    <property name="spanCollector" ref="scribe"/>
  </bean>

ServiceName

However if you want to map Camel endpoints to human friendly logical names, you can add mappings

• ServiceName
  
  

You can configure a global service name that all events will fallback and use, such as:

zipkin.setServiceName("invoices");

This will use the same service name for all incoming and outgoing Zipkin traces. So if your application uses different services, you need to map them more fine grained into client vs server mappings

Client and Server Service Mappings

• ClientServiceMappings
• ServerServiceMappings

So if your application hosts a service that others can call, you can map the Camel route endpoint to a server service mapping. For example support your Camel application has the following route

from("activemq:queue:inbox")
  ...
  .to("http:someserver/somepath");

And you want to make that as a server service, you can add the following mapping

zipkin.addServerServiceMapping("activemq:queue:inbox", "orders");

Then when a message is consumed from that inbox queue, it becomes a Zipkin server event with the service name orders.

Now suppose that the call to http:someserver/somepath is also a service, which you want to map to a client service name, which can be done as:

zipkin.addClientServiceMapping("http:someserver/somepath", "audit");

Then in the same Camel application you have mapped incoming and outgoing endpoints to different zipkin service names.

zipkin.addClientServiceMapping("http:someserver*", "audit");

Mapping rules

The service name mapping for server occurs using the following rules

1. Is there an exclude pattern that matches the endpoint URI of the from endpoint? If yes then skip.
2. Is there a match in the serviceServiceMapping that matches the endpoint URI of the from endpoint? If yes the use the found service name
3. Is there a match in the serviceServiceMapping that matches the route id of the current route? If yes the use the found service name
4. Is there a match in the serviceServiceMapping that matches the original route id where the exchange started? If yes the use the found service name
5. No service name was found, the exchange is not traced by Zipkin.

The service name mapping for client occurs using the following rules

1. Is there an exclude pattern that matches the endpoint URI of the from endpoint? If yes then skip.
2. Is there a match in the clientServiceMapping that matches the endpoint URI of endpoint where the message is being sent to? If yes the use the found service name
3. Is there a match in the clientServiceMapping that matches the route id of the current route? If yes the use the found service name
4. Is there a match in the clientServiceMapping that matches the original route id where the exchange started? If yes the use the found service name
5. No service name was found, the exchange is not traced by Zipkin.

No client or server mappings

If there has been no configuration of client or server service mappings, then CamelZipkin runs in a fallback mode, where it uses the endpoint URIs as the service name.

So in the example above that would mean the service names would be, as if you add the following code yourself:

zipkin.addServerServiceMapping("activemq:queue:inbox", "activemq:queue:inbox");
zipkin.addClientServiceMapping("http:someserver/somepath", "http:someserver/somepath");

This is not a recommended approach but gets you up and running quickly without doing any service name mappings. However when you have multiple systems across your infrastructure, then you should consider using human logic service names, that you map to instead of using the camel endpoint URIs.

camel-zipkin-starter

If you are using Spring Boot then you can add the camel-zipkin-starter dependency, and turn on Zipkin by annotating the main class with @CamelZipkin. You can then configure camel-zipkin in the application.properties file where you can configure the hostname and port number for the Zipkin Server, and all the other options as listed in the options table above.

You can find an example of this in the camel-example-zipkin