nameko-zipkin

Zipkin tracing for nameko framework

Install

pip install nameko-zipkin 

Usage

Services

from nameko_zipkin import Zipkin
from nameko.rpc import rpc


class Service:
    name = 'service'
    zipkin = Zipkin() # Dependency provider injects py_zipkin.zipkin.zipkin_span object
    
    @rpc
    def method(self):
        assert self.zipkin.service_name == Service.name
        assert self.zipkin.span_name == Service.method.__name__

Standalone rpc

from py_zipkin import zipkin
from nameko_zipkin.transport import HttpHandler
from nameko.standalone.rpc import ClusterRpcProxy


handler = HttpHandler('http://localhost:9411/api/v1/spans').handle

with zipkin.zipkin_server_span('RootService',
                               'RootMethod',
                               sample_rate=100.,
                               transport_handler=handler):
    with ClusterRpcProxy({
        'AMQP_URI': "pyamqp://guest:guest@localhost"
    }, context_data=zipkin.create_http_headers_for_new_span()) as proxy:
        proxy.service.method()

How it works

• On service method call new server span is created
• Trace parameters (trace_id, parent_span_id, etc.) are captured from context data and are accessible in py_zipkin.thread_local.get_zipkin_attrs
• If there are no parameters, request isn't traced
• Child service calls are also supported
• Trace results are reported through handler classes in nameko_zipkin.transport

Note: Span is initiated on server so only sr and ss annotations are reported, that is not the proper way. It should be initiated on client but this approach requires changes in nameko framework.

Configuration

ZIPKIN section must be added to nameko service config.yaml

ZIPKIN:
    HANDLER: HttpHandler
    HANDLER_PARAMS:
      url: http://localhost:9411/api/v1/spans

Planed changes

• Trace initialization on trace attrs absence in context data
• Kafka transport support
• Custom handlers support in config.yaml ('my_module.MyHandler')