Nameko¶
[nah-meh-koh]
A microservices framework for Python that lets service developers concentrate on application logic and encourages testability.
A nameko service is just a class:
# helloworld.py
from nameko.rpc import rpc
class GreetingService:
name = "greeting_service"
@rpc
def hello(self, name):
return "Hello, {}!".format(name)
Note
The example above requires RabbitMQ, because it’s using the built-in AMQP RPC features. RabbitMQ installation guidelines offer several installation options, but you can quickly install and run it using Docker.
To install and run RabbitMQ using docker:
$ docker run -d -p 5672:5672 rabbitmq:3
You can run it in a shell:
$ nameko run helloworld
starting services: greeting_service
...
And play with it from another:
$ nameko shell
>>> n.rpc.greeting_service.hello(name="ナメコ")
'Hello, ナメコ!'
User Guide¶
This section covers most things you need to know to create and run your own Nameko services.
What is Nameko?¶
Nameko is a framework for building microservices in Python.
It comes with built-in support for:
- RPC over AMQP
- Asynchronous events (pub-sub) over AMQP
- Simple HTTP GET and POST
- Websocket RPC and subscriptions (experimental)
Out of the box you can build a service that can respond to RPC messages, dispatch events on certain actions, and listen to events from other services. It could also have HTTP interfaces for clients that can’t speak AMQP, and a websocket interface for, say, Javascript clients.
Nameko is also extensible. You can define your own transport mechanisms and service dependencies to mix and match as desired.
Nameko strongly encourages the dependency injection pattern, which makes building and testing services clean and simple.
Nameko takes its name from the Japanese mushroom, which grows in clusters.
When should I use Nameko?¶
Nameko is designed to help you create, run and test microservices. You should use Nameko if:
- You want to write your backend as microservices, or
- You want to add microservices to an existing system, and
- You want to do it in Python.
Nameko scales from a single instance of a single service, to a cluster with many instances of many different services.
The library also ships with tools for clients, which you can use if you want to write Python code to communicate with an existing Nameko cluster.
Key Concepts¶
This section introduces Nameko’s central concepts.
Anatomy of a Service¶
A Nameko service is just a Python class. The class encapsulates the application logic in its methods and declares any dependencies as attributes.
Methods are exposed to the outside world with entrypoint decorators.
from nameko.rpc import rpc, RpcProxy
class Service:
name = "service"
# we depend on the RPC interface of "another_service"
other_rpc = RpcProxy("another_service")
@rpc # `method` is exposed over RPC
def method(self):
# application logic goes here
pass
Entrypoints¶
Entrypoints are gateways into the service methods they decorate. They normally monitor an external entity, for example a message queue. On a relevant event, the entrypoint may “fire” and the decorated method would be executed by a service worker.
Dependencies¶
Most services depend on something other than themselves. Nameko encourages these things to be implemented as dependencies.
A dependency is an opportunity to hide code that isn’t part of the core service logic. The dependency’s interface to the service should be as simple as possible.
Declaring dependencies in your service is a good idea for lots of reasons, and you should think of them as the gateway between service code and everything else. That includes other services, external APIs, and even databases.
Workers¶
Workers are created when an entrypoint fires. A worker is just an instance of the service class, but with the dependency declarations replaced with instances of those dependencies (see dependency injection).
Note that a worker only lives for the execution of one method – services are stateless from one call to the next, which encourages the use of dependencies.
A service can run multiple workers at the same time, up to a user-defined limit. See concurrency for details.
Dependency Injection¶
Adding a dependency to a service class is declarative. That is, the attribute on the class is a declaration, rather than the interface that workers can actually use.
The class attribute is a DependencyProvider
. It is responsible for providing an object that is injected into service workers.
Dependency providers implement a get_dependency()
method, the result of which is injected into a newly created worker.
The lifecycle of a worker is:
- Entrypoint fires
- Worker instantiated from service class
- Dependencies injected into worker
- Method executes
- Worker is destroyed
In pseudocode this looks like:
worker = Service()
worker.other_rpc = worker.other_rpc.get_dependency()
worker.method()
del worker
Dependency providers live for the duration of the service, whereas the injected dependency can be unique to each worker.
Concurrency¶
Nameko is built on top of the eventlet library, which provides concurrency via “greenthreads”. The concurrency model is co-routines with implicit yielding.
Implicit yielding relies on monkey patching the standard library, to trigger a yield when a thread waits on I/O. If you host services with nameko run
on the command line, Nameko will apply the monkey patch for you.
Each worker executes in its own greenthread. The maximum number of concurrent workers can be tweaked based on the amount of time each worker will spend waiting on I/O.
Workers are stateless so are inherently thread safe, but dependencies should ensure they are unique per worker or otherwise safe to be accessed concurrently by multiple workers.
Note that many C-extensions that are using sockets and that would normally be considered thread-safe may not work with greenthreads. Among them are librabbitmq, MySQLdb and others.
Extensions¶
All entrypoints and dependency providers are implemented as “extensions”. We refer to them this way because they’re outside of service code but are not required by all services (for example, a purely AMQP-exposed service won’t use the HTTP entrypoints).
Nameko has a number of built-in extensions, some are provided by the community and you can write your own.
Running Services¶
All that’s required to run a service is the service class and any relevant configuration. The easiest way to run one or multiple services is with the Nameko CLI:
$ nameko run module:[ServiceClass]
This command will discover Nameko services in the given module
s and start running them. You can optionally limit it to specific ServiceClass
s.
Service Containers¶
Each service class is delegated to a ServiceContainer
. The container encapsulates all the functionality required to run a service, and also encloses any extensions on the service class.
Using the ServiceContainer
to run a single service:
from nameko.containers import ServiceContainer
class Service:
name = "service"
# create a container
container = ServiceContainer(Service, config={})
# ``container.extensions`` exposes all extensions used by the service
service_extensions = list(container.extensions)
# start service
container.start()
# stop service
container.stop()
Service Runner¶
ServiceRunner
is a thin wrapper around multiple containers, exposing methods for starting and stopping all the wrapped containers simultaneously. This is what nameko run
uses internally, and it can also be constructed programmatically:
from nameko.runners import ServiceRunner
from nameko.testing.utils import get_container
class ServiceA:
name = "service_a"
class ServiceB:
name = "service_b"
# create a runner for ServiceA and ServiceB
runner = ServiceRunner(config={})
runner.add_service(ServiceA)
runner.add_service(ServiceB)
# ``get_container`` will return the container for a particular service
container_a = get_container(runner, ServiceA)
# start both services
runner.start()
# stop both services
runner.stop()
If you create your own runner rather than using nameko run, you must also apply the eventlet monkey patch.
See the nameko.cli.run module for an example.
Installation¶
Install with Pip¶
You can install nameko and its dependencies from PyPI with pip:
pip install nameko
Source Code¶
Nameko is actively developed on GitHub. Get the code by cloning the public repository:
git clone git@github.com:nameko/nameko.git
You can install from the source code using setuptools:
python setup.py install
RabbitMQ¶
Several of Nameko’s built-in features rely on RabbitMQ. Installing RabbitMQ is straightforward on most platforms and they have excellent documentation.
With homebrew on a mac you can install with:
brew install rabbitmq
On debian-based operating systems:
apt-get install rabbitmq-server
For other platforms, consult the RabbitMQ installation guidelines.
The RabbitMQ broker will be ready to go as soon as it’s installed – it doesn’t need any configuration. The examples in this documentation assume you have a broker running on the default ports on localhost and the rabbitmq_management plugin is enabled.
Command Line Interface¶
Nameko ships with a command line interface to make hosting and interacting with services as easy as possible.
Running a Service¶
$ nameko run <module>[:<ServiceClass>]
Discover and run a service class. This will start the service in the foreground and run it until the process terminates.
It is possible to override the default settings using a --config
switch
$ nameko run --config ./foobar.yaml <module>[:<ServiceClass>]
and providing a simple YAML configuration file:
# foobar.yaml
AMQP_URI: 'pyamqp://guest:guest@localhost'
WEB_SERVER_ADDRESS: '0.0.0.0:8000'
rpc_exchange: 'nameko-rpc'
max_workers: 10
parent_calls_tracked: 10
LOGGING:
version: 1
handlers:
console:
class: logging.StreamHandler
root:
level: DEBUG
handlers: [console]
The LOGGING
entry is passed to logging.config.dictConfig()
and should conform to the schema for that call.
Config values can be read via the built-in Config dependency provider.
Environment variable substitution¶
YAML configuration files have support for environment variables.
You can use bash style syntax: ${ENV_VAR}
.
Optionally you can provide default values ${ENV_VAR:default_value}
.
Default values can contains environment variables recursively ${ENV_VAR:default_${OTHER_ENV_VAR:value}}
(note: this feature requires regex package).
# foobar.yaml
AMQP_URI: pyamqp://${RABBITMQ_USER:guest}:${RABBITMQ_PASSWORD:password}@${RABBITMQ_HOST:localhost}
To run your service and set environment variables for it to use:
$ RABBITMQ_USER=user RABBITMQ_PASSWORD=password RABBITMQ_HOST=host nameko run --config ./foobar.yaml <module>[:<ServiceClass>]
If you need to quote the values in your YAML file, the explicit !env_var
resolver is required:
# foobar.yaml
AMQP_URI: !env_var "pyamqp://${RABBITMQ_USER:guest}:${RABBITMQ_PASSWORD:password}@${RABBITMQ_HOST:localhost}"
If you need to use values as raw strings in your YAML file without them getting converted to native python,
the explicit !raw_env_var
resolver is required:
# foobar.yaml
ENV_THAT_IS_NEEDED_RAW: !raw_env_var "${ENV_THAT_IS_NEEDED_RAW:1234.5660}"
This will turn into the string value 1234.5660
, instead of a float number.
You can provide many levels of default values
# foobar.yaml
AMQP_URI: ${AMQP_URI:pyamqp://${RABBITMQ_USER:guest}:${RABBITMQ_PASSWORD:password}@${RABBITMQ_HOST:localhost}}
this config accepts AMQP_URI as an environment variable, if provided RABBITMQ_* nested variables will not be used.
The environment variable value is interpreted as YAML, so it is possible to use rich types:
# foobar.yaml
...
THINGS: ${A_LIST_OF_THINGS}
$ A_LIST_OF_THINGS=[A,B,C] nameko run --config ./foobar.yaml <module>[:<ServiceClass>]
the parser for environment variables will pair all brackets.
# foobar.yaml
LANDING_URL_TEMPLATE: ${LANDING_URL_TEMPLATE:https://example.com/{path}}
so the default value for this config will be https://example.com/{path}
Interacting with running services¶
$ nameko shell
Launch an interactive python shell for working with remote nameko services. This is a regular interactive interpreter, with a special module n
added
to the built-in namespace, providing the ability to make RPC calls and dispatch events.
Making an RPC call to “target_service”:
$ nameko shell
>>> n.rpc.target_service.target_method(...)
# RPC response
Dispatching an event as “source_service”:
$ nameko shell
>>> n.dispatch_event("source_service", "event_type", "event_payload")
Built-in Extensions¶
Nameko includes a number of built-in extensions. This section introduces them and gives brief examples of their usage.
RPC¶
Nameko includes an implementation of RPC over AMQP. It comprises the @rpc
entrypoint, a proxy for services to talk to other services, and a standalone proxy that non-Nameko clients can use to make RPC calls to a cluster:
from nameko.rpc import rpc, RpcProxy
class ServiceY:
name = "service_y"
@rpc
def append_identifier(self, value):
return u"{}-y".format(value)
class ServiceX:
name = "service_x"
y = RpcProxy("service_y")
@rpc
def remote_method(self, value):
res = u"{}-x".format(value)
return self.y.append_identifier(res)
from nameko.standalone.rpc import ClusterRpcProxy
config = {
'AMQP_URI': AMQP_URI # e.g. "pyamqp://guest:guest@localhost"
}
with ClusterRpcProxy(config) as cluster_rpc:
cluster_rpc.service_x.remote_method("hellø") # "hellø-x-y"
Normal RPC calls block until the remote method completes, but proxies also have an asynchronous calling mode to background or parallelize RPC calls:
with ClusterRpcProxy(config) as cluster_rpc:
hello_res = cluster_rpc.service_x.remote_method.call_async("hello")
world_res = cluster_rpc.service_x.remote_method.call_async("world")
# do work while waiting
hello_res.result() # "hello-x-y"
world_res.result() # "world-x-y"
In a cluster with more than one instance of the target service, RPC requests round-robin between instances. The request will be handled by exactly one instance of the target service.
AMQP messages are ack’d only after the request has been successfully processed. If the service fails to acknowledge the message and the AMQP connection is closed (e.g. if the service process is killed) the broker will revoke and then allocate the message to the available service instance.
Request and response payloads are serialized into JSON for transport over the wire.
Events (Pub-Sub)¶
Nameko Events is an asynchronous messaging system, implementing the Publish-Subscriber pattern. Services dispatch events that may be received by zero or more others:
from nameko.events import EventDispatcher, event_handler
from nameko.rpc import rpc
class ServiceA:
""" Event dispatching service. """
name = "service_a"
dispatch = EventDispatcher()
@rpc
def dispatching_method(self, payload):
self.dispatch("event_type", payload)
class ServiceB:
""" Event listening service. """
name = "service_b"
@event_handler("service_a", "event_type")
def handle_event(self, payload):
print("service b received:", payload)
The EventHandler
entrypoint has three handler_type
s that determine how event messages are received in a cluster:
SERVICE_POOL
– event handlers are pooled by service name and one instance from each pool receives the event, similar to the cluster behaviour of the RPC entrypoint. This is the default handler type.BROADCAST
– every listening service instance will receive the event.SINGLETON
– exactly one listening service instance will receive the event.
An example of using the BROADCAST
mode:
from nameko.events import BROADCAST, event_handler
class ListenerService:
name = "listener"
@event_handler(
"monitor", "ping", handler_type=BROADCAST, reliable_delivery=False
)
def ping(self, payload):
# all running services will respond
print("pong from {}".format(self.name))
Events are serialized into JSON for transport over the wire.
HTTP¶
The HTTP entrypoint is built on top of werkzeug, and supports all the standard HTTP methods (GET/POST/DELETE/PUT etc)
The HTTP entrypoint can specify multiple HTTP methods for a single URL as a comma-separated list. See example below.
Service methods must return one of:
- a string, which becomes the response body
- a 2-tuple
(status code, response body)
- a 3-tuple
(status_code, headers dict, response body)
- an instance of
werkzeug.wrappers.Response
# http.py
import json
from nameko.web.handlers import http
class HttpService:
name = "http_service"
@http('GET', '/get/<int:value>')
def get_method(self, request, value):
return json.dumps({'value': value})
@http('POST', '/post')
def do_post(self, request):
return u"received: {}".format(request.get_data(as_text=True))
@http('GET,PUT,POST,DELETE', '/multi')
def do_multi(self, request):
return request.method
$ nameko run http
starting services: http_service
$ curl -i localhost:8000/get/42
HTTP/1.1 200 OK
Content-Type: text/plain; charset=utf-8
Content-Length: 13
Date: Fri, 13 Feb 2015 14:51:18 GMT
{'value': 42}
$ curl -i -d "post body" localhost:8000/post
HTTP/1.1 200 OK
Content-Type: text/plain; charset=utf-8
Content-Length: 19
Date: Fri, 13 Feb 2015 14:55:01 GMT
received: post body
A more advanced example:
# advanced_http.py
from nameko.web.handlers import http
from werkzeug.wrappers import Response
class Service:
name = "advanced_http_service"
@http('GET', '/privileged')
def forbidden(self, request):
return 403, "Forbidden"
@http('GET', '/headers')
def redirect(self, request):
return 201, {'Location': 'https://www.example.com/widget/1'}, ""
@http('GET', '/custom')
def custom(self, request):
return Response("payload")
$ nameko run advanced_http
starting services: advanced_http_service
$ curl -i localhost:8000/privileged
HTTP/1.1 403 FORBIDDEN
Content-Type: text/plain; charset=utf-8
Content-Length: 9
Date: Fri, 13 Feb 2015 14:58:02 GMT
curl -i localhost:8000/headers
HTTP/1.1 201 CREATED
Location: https://www.example.com/widget/1
Content-Type: text/plain; charset=utf-8
Content-Length: 0
Date: Fri, 13 Feb 2015 14:58:48 GMT
You can control formatting of errors returned from your service by overriding response_from_exception()
:
import json
from nameko.web.handlers import HttpRequestHandler
from werkzeug.wrappers import Response
from nameko.exceptions import safe_for_serialization
class HttpError(Exception):
error_code = 'BAD_REQUEST'
status_code = 400
class InvalidArgumentsError(HttpError):
error_code = 'INVALID_ARGUMENTS'
class HttpEntrypoint(HttpRequestHandler):
def response_from_exception(self, exc):
if isinstance(exc, HttpError):
response = Response(
json.dumps({
'error': exc.error_code,
'message': safe_for_serialization(exc),
}),
status=exc.status_code,
mimetype='application/json'
)
return response
return HttpRequestHandler.response_from_exception(self, exc)
http = HttpEntrypoint.decorator
class Service:
name = "http_service"
@http('GET', '/custom_exception')
def custom_exception(self, request):
raise InvalidArgumentsError("Argument `foo` is required.")
$ nameko run http_exceptions
starting services: http_service
$ curl -i http://localhost:8000/custom_exception
HTTP/1.1 400 BAD REQUEST
Content-Type: application/json
Content-Length: 72
Date: Thu, 06 Aug 2015 09:53:56 GMT
{"message": "Argument `foo` is required.", "error": "INVALID_ARGUMENTS"}
You can change the HTTP port and IP using the WEB_SERVER_ADDRESS config setting:
# foobar.yaml
AMQP_URI: 'pyamqp://guest:guest@localhost'
WEB_SERVER_ADDRESS: '0.0.0.0:8000'
Timer¶
The Timer
is a simple entrypoint that fires once per a configurable number of seconds. The timer is not “cluster-aware” and fires on all services instances.
from nameko.timer import timer
class Service:
name ="service"
@timer(interval=1)
def ping(self):
# method executed every second
print("pong")
Built-in Dependency Providers¶
Nameko includes some commonly used dependency providers. This section introduces them and gives brief examples of their usage.
Config¶
Config is a simple dependency provider that gives services read-only access to configuration values at run time, see Running a Service.
from nameko.dependency_providers import Config
from nameko.web.handlers import http
class Service:
name = "test_config"
config = Config()
@property
def foo_enabled(self):
return self.config.get('FOO_FEATURE_ENABLED', False)
@http('GET', '/foo')
def foo(self, request):
if not self.foo_enabled:
return 403, "FeatureNotEnabled"
return 'foo'
Community¶
There are a number of nameko extensions and supplementary libraries that are not part of the core project but that you may find useful when developing your own nameko services:
Extensions¶
A
DependencyProvider
for writing to databases with SQLAlchemy. Requires a pure-python or otherwise eventlet-compatible database driver.Consider combining it with SQLAlchemy-filters to add filtering, sorting and pagination of query objects when exposing them over a REST API.
Captures entrypoint exceptions and sends tracebacks to a Sentry server.
Nameko extension allowing AMQP entrypoints to retry later.
Nameko extension with a Cometd client implementing Bayeux protocol
Nameko extension for interaction with Slack APIs. Uses Slack Developer Kit for Python.
Nameko dependency provider that dispatches log data using Events (Pub-Sub).
Redis dependency and utils for Nameko.
Redis dependency for nameko services
A StatsD dependency for nameko, enabling services to send stats.
Twilio dependency for nameko, so you can send SMS, make calls, and answer calls in your service.
SendGrid dependency for nameko, for sending transactional and marketing emails.
Tools to cache RPC interactions between your nameko services.
Supplementary Libraries¶
Django wrapper for Nameko microservice framework.
A wrapper for using nameko services with Flask.
Standalone async proxy to communicate with Nameko microservices.
Search PyPi for more nameko packages
If you would like your own nameko extension or library to appear on this page, please get in touch.
Testing Services¶
Philosophy¶
Nameko’s conventions are designed to make testing as easy as possible. Services are likely to be small and single-purpose, and dependency injection makes it simple to replace and isolate pieces of functionality.
The examples below use pytest, which is what Nameko’s own test suite uses, but the helpers are test framework agnostic.
Unit Testing¶
Unit testing in Nameko usually means testing a single service in isolation – i.e. without any or most of its dependencies.
The worker_factory()
utility will create a worker from a given service class, with its dependencies replaced by mock.MagicMock
objects. Dependency functionality can then be imitated by adding side_effect
s and return_value
s:
""" Service unit testing best practice.
"""
from nameko.rpc import RpcProxy, rpc
from nameko.testing.services import worker_factory
class ConversionService(object):
""" Service under test
"""
name = "conversions"
maths_rpc = RpcProxy("maths")
@rpc
def inches_to_cm(self, inches):
return self.maths_rpc.multiply(inches, 2.54)
@rpc
def cms_to_inches(self, cms):
return self.maths_rpc.divide(cms, 2.54)
def test_conversion_service():
# create worker with mock dependencies
service = worker_factory(ConversionService)
# add side effects to the mock proxy to the "maths" service
service.maths_rpc.multiply.side_effect = lambda x, y: x * y
service.maths_rpc.divide.side_effect = lambda x, y: x / y
# test inches_to_cm business logic
assert service.inches_to_cm(300) == 762
service.maths_rpc.multiply.assert_called_once_with(300, 2.54)
# test cms_to_inches business logic
assert service.cms_to_inches(762) == 300
service.maths_rpc.divide.assert_called_once_with(762, 2.54)
In some circumstances it’s helpful to provide an alternative dependency, rather than use a mock. This may be a fully functioning replacement (e.g. a test database session) or a lightweight shim that provides partial functionality.
""" Service unit testing best practice, with an alternative dependency.
"""
import pytest
from sqlalchemy import Column, Integer, String, create_engine
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import sessionmaker
from nameko.rpc import rpc
from nameko.testing.services import worker_factory
# using community extension from http://pypi.python.org/pypi/nameko-sqlalchemy
from nameko_sqlalchemy import Session
Base = declarative_base()
class Result(Base):
__tablename__ = 'model'
id = Column(Integer, primary_key=True)
value = Column(String(64))
class Service:
""" Service under test
"""
name = "service"
db = Session(Base)
@rpc
def save(self, value):
result = Result(value=value)
self.db.add(result)
self.db.commit()
@pytest.fixture
def session():
""" Create a test database and session
"""
engine = create_engine('sqlite:///:memory:')
Base.metadata.create_all(engine)
session_cls = sessionmaker(bind=engine)
return session_cls()
def test_service(session):
# create instance, providing the test database session
service = worker_factory(Service, db=session)
# verify ``save`` logic by querying the test database
service.save("helloworld")
assert session.query(Result.value).all() == [("helloworld",)]
Integration Testing¶
Integration testing in Nameko means testing the interface between a number of services. The recommended way is to run all the services being tested in the normal way, and trigger behaviour by “firing” an entrypoint using a helper:
""" Service integration testing best practice.
"""
from nameko.rpc import rpc, RpcProxy
from nameko.testing.utils import get_container
from nameko.testing.services import entrypoint_hook
class ServiceX:
""" Service under test
"""
name = "service_x"
y = RpcProxy("service_y")
@rpc
def remote_method(self, value):
res = "{}-x".format(value)
return self.y.append_identifier(res)
class ServiceY:
""" Service under test
"""
name = "service_y"
@rpc
def append_identifier(self, value):
return "{}-y".format(value)
def test_service_x_y_integration(runner_factory, rabbit_config):
# run services in the normal manner
runner = runner_factory(rabbit_config, ServiceX, ServiceY)
runner.start()
# artificially fire the "remote_method" entrypoint on ServiceX
# and verify response
container = get_container(runner, ServiceX)
with entrypoint_hook(container, "remote_method") as entrypoint:
assert entrypoint("value") == "value-x-y"
Note that the interface between ServiceX
and ServiceY
here is just as if under normal operation.
Interfaces that are out of scope for a particular test can be deactivated with one of the following test helpers:
restrict_entrypoints¶
-
nameko.testing.services.
restrict_entrypoints
(container, *entrypoints) Restrict the entrypoints on
container
to those named inentrypoints
.This method must be called before the container is started.
Usage
The following service definition has two entrypoints:
class Service(object): name = "service" @timer(interval=1) def foo(self, arg): pass @rpc def bar(self, arg) pass @rpc def baz(self, arg): pass container = ServiceContainer(Service, config)
To disable the timer entrypoint on
foo
, leaving just the RPC entrypoints:restrict_entrypoints(container, "bar", "baz")
Note that it is not possible to identify multiple entrypoints on the same method individually.
replace_dependencies¶
-
nameko.testing.services.
replace_dependencies
(container, *dependencies, **dependency_map) Replace the dependency providers on
container
with instances ofMockDependencyProvider
.Dependencies named in *dependencies will be replaced with a
MockDependencyProvider
, which injects a MagicMock instead of the dependency.Alternatively, you may use keyword arguments to name a dependency and provide the replacement value that the MockDependencyProvider should inject.
Return the
MockDependencyProvider.dependency
for every dependency specified in the (*dependencies) args so that calls to the replaced dependencies can be inspected. Return a single object if only one dependency was replaced, and a generator yielding the replacements in the same order asdependencies
otherwise. Note that any replaced dependencies specified via kwargs **dependency_map will not be returned.Replacements are made on the container instance and have no effect on the service class. New container instances are therefore unaffected by replacements on previous instances.
Usage
from nameko.rpc import RpcProxy, rpc from nameko.standalone.rpc import ServiceRpcProxy class ConversionService(object): name = "conversions" maths_rpc = RpcProxy("maths") @rpc def inches_to_cm(self, inches): return self.maths_rpc.multiply(inches, 2.54) @rpc def cm_to_inches(self, cms): return self.maths_rpc.divide(cms, 2.54) container = ServiceContainer(ConversionService, config) mock_maths_rpc = replace_dependencies(container, "maths_rpc") mock_maths_rpc.divide.return_value = 39.37 container.start() with ServiceRpcProxy('conversions', config) as proxy: proxy.cm_to_inches(100) # assert that the dependency was called as expected mock_maths_rpc.divide.assert_called_once_with(100, 2.54)
Providing a specific replacement by keyword:
class StubMaths(object): def divide(self, val1, val2): return val1 / val2 replace_dependencies(container, maths_rpc=StubMaths()) container.start() with ServiceRpcProxy('conversions', config) as proxy: assert proxy.cm_to_inches(127) == 50.0
Complete Example¶
The following integration testing example makes use of both scope-restricting helpers:
"""
This file defines several toy services that interact to form a shop of the
famous ACME Corporation. The AcmeShopService relies on the StockService,
InvoiceService and PaymentService to fulfil its orders. They are not best
practice examples! They're minimal services provided for the test at the
bottom of the file.
``test_shop_integration`` is a full integration test of the ACME shop
"checkout flow". It demonstrates how to test the multiple ACME services in
combination with each other, including limiting service interactions by
replacing certain entrypoints and dependencies.
"""
from collections import defaultdict
import pytest
from nameko.extensions import DependencyProvider
from nameko.events import EventDispatcher, event_handler
from nameko.exceptions import RemoteError
from nameko.rpc import rpc, RpcProxy
from nameko.standalone.rpc import ServiceRpcProxy
from nameko.testing.services import replace_dependencies, restrict_entrypoints
from nameko.testing.utils import get_container
from nameko.timer import timer
class NotLoggedInError(Exception):
pass
class ItemOutOfStockError(Exception):
pass
class ItemDoesNotExistError(Exception):
pass
class ShoppingBasket(DependencyProvider):
""" A shopping basket tied to the current ``user_id``.
"""
def __init__(self):
self.baskets = defaultdict(list)
def get_dependency(self, worker_ctx):
class Basket(object):
def __init__(self, basket):
self._basket = basket
self.worker_ctx = worker_ctx
def add(self, item):
self._basket.append(item)
def __iter__(self):
for item in self._basket:
yield item
try:
user_id = worker_ctx.data['user_id']
except KeyError:
raise NotLoggedInError()
return Basket(self.baskets[user_id])
class AcmeShopService:
name = "acmeshopservice"
user_basket = ShoppingBasket()
stock_rpc = RpcProxy('stockservice')
invoice_rpc = RpcProxy('invoiceservice')
payment_rpc = RpcProxy('paymentservice')
fire_event = EventDispatcher()
@rpc
def add_to_basket(self, item_code):
""" Add item identified by ``item_code`` to the shopping basket.
This is a toy example! Ignore the obvious race condition.
"""
stock_level = self.stock_rpc.check_stock(item_code)
if stock_level > 0:
self.user_basket.add(item_code)
self.fire_event("item_added_to_basket", item_code)
return item_code
raise ItemOutOfStockError(item_code)
@rpc
def checkout(self):
""" Take payment for all items in the shopping basket.
"""
total_price = sum(self.stock_rpc.check_price(item)
for item in self.user_basket)
# prepare invoice
invoice = self.invoice_rpc.prepare_invoice(total_price)
# take payment
self.payment_rpc.take_payment(invoice)
# fire checkout event if prepare_invoice and take_payment succeeded
checkout_event_data = {
'invoice': invoice,
'items': list(self.user_basket)
}
self.fire_event("checkout_complete", checkout_event_data)
return total_price
class Warehouse(DependencyProvider):
""" A database of items in the warehouse.
This is a toy example! A dictionary is not a database.
"""
def __init__(self):
self.database = {
'anvil': {
'price': 100,
'stock': 3
},
'dehydrated_boulders': {
'price': 999,
'stock': 12
},
'invisible_paint': {
'price': 10,
'stock': 30
},
'toothpicks': {
'price': 1,
'stock': 0
}
}
def get_dependency(self, worker_ctx):
return self.database
class StockService:
name = "stockservice"
warehouse = Warehouse()
@rpc
def check_price(self, item_code):
""" Check the price of an item.
"""
try:
return self.warehouse[item_code]['price']
except KeyError:
raise ItemDoesNotExistError(item_code)
@rpc
def check_stock(self, item_code):
""" Check the stock level of an item.
"""
try:
return self.warehouse[item_code]['stock']
except KeyError:
raise ItemDoesNotExistError(item_code)
@rpc
@timer(100)
def monitor_stock(self):
""" Periodic stock monitoring method. Can also be triggered manually
over RPC.
This is an expensive process that we don't want to exercise during
integration testing...
"""
raise NotImplementedError()
@event_handler('acmeshopservice', "checkout_complete")
def dispatch_items(self, event_data):
""" Dispatch items from stock on successful checkouts.
This is an expensive process that we don't want to exercise during
integration testing...
"""
raise NotImplementedError()
class AddressBook(DependencyProvider):
""" A database of user details, keyed on user_id.
"""
def __init__(self):
self.address_book = {
'wile_e_coyote': {
'username': 'wile_e_coyote',
'fullname': 'Wile E Coyote',
'address': '12 Long Road, High Cliffs, Utah',
},
}
def get_dependency(self, worker_ctx):
def get_user_details():
try:
user_id = worker_ctx.data['user_id']
except KeyError:
raise NotLoggedInError()
return self.address_book.get(user_id)
return get_user_details
class InvoiceService:
name = "invoiceservice"
get_user_details = AddressBook()
@rpc
def prepare_invoice(self, amount):
""" Prepare an invoice for ``amount`` for the current user.
"""
address = self.get_user_details().get('address')
fullname = self.get_user_details().get('fullname')
username = self.get_user_details().get('username')
msg = "Dear {}. Please pay ${} to ACME Corp.".format(fullname, amount)
invoice = {
'message': msg,
'amount': amount,
'customer': username,
'address': address
}
return invoice
class PaymentService:
name = "paymentservice"
@rpc
def take_payment(self, invoice):
""" Take payment from a customer according to ``invoice``.
This is an expensive process that we don't want to exercise during
integration testing...
"""
raise NotImplementedError()
# =============================================================================
# Begin test
# =============================================================================
@pytest.yield_fixture
def rpc_proxy_factory(rabbit_config):
""" Factory fixture for standalone RPC proxies.
Proxies are started automatically so they can be used without a ``with``
statement. All created proxies are stopped at the end of the test, when
this fixture closes.
"""
all_proxies = []
def make_proxy(service_name, **kwargs):
proxy = ServiceRpcProxy(service_name, rabbit_config, **kwargs)
all_proxies.append(proxy)
return proxy.start()
yield make_proxy
for proxy in all_proxies:
proxy.stop()
def test_shop_checkout_integration(
rabbit_config, runner_factory, rpc_proxy_factory
):
""" Simulate a checkout flow as an integration test.
Requires instances of AcmeShopService, StockService and InvoiceService
to be running. Explicitly replaces the rpc proxy to PaymentService so
that service doesn't need to be hosted.
Also replaces the event dispatcher dependency on AcmeShopService and
disables the timer entrypoint on StockService. Limiting the interactions
of services in this way reduces the scope of the integration test and
eliminates undesirable side-effects (e.g. processing events unnecessarily).
"""
context_data = {'user_id': 'wile_e_coyote'}
shop = rpc_proxy_factory('acmeshopservice', context_data=context_data)
runner = runner_factory(
rabbit_config, AcmeShopService, StockService, InvoiceService)
# replace ``event_dispatcher`` and ``payment_rpc`` dependencies on
# AcmeShopService with ``MockDependencyProvider``\s
shop_container = get_container(runner, AcmeShopService)
fire_event, payment_rpc = replace_dependencies(
shop_container, "fire_event", "payment_rpc")
# restrict entrypoints on StockService
stock_container = get_container(runner, StockService)
restrict_entrypoints(stock_container, "check_price", "check_stock")
runner.start()
# add some items to the basket
assert shop.add_to_basket("anvil") == "anvil"
assert shop.add_to_basket("invisible_paint") == "invisible_paint"
# try to buy something that's out of stock
with pytest.raises(RemoteError) as exc_info:
shop.add_to_basket("toothpicks")
assert exc_info.value.exc_type == "ItemOutOfStockError"
# provide a mock response from the payment service
payment_rpc.take_payment.return_value = "Payment complete."
# checkout
res = shop.checkout()
total_amount = 100 + 10
assert res == total_amount
# verify integration with mocked out payment service
payment_rpc.take_payment.assert_called_once_with({
'customer': "wile_e_coyote",
'address': "12 Long Road, High Cliffs, Utah",
'amount': total_amount,
'message': "Dear Wile E Coyote. Please pay $110 to ACME Corp."
})
# verify events fired as expected
assert fire_event.call_count == 3
if __name__ == "__main__":
import sys
pytest.main(sys.argv)
Other Helpers¶
entrypoint_hook¶
The entrypoint hook allows a service entrypoint to be called manually. This is useful during integration testing if it is difficult or expensive to fake to external event that would cause an entrypoint to fire.
You can provide context_data for the call to mimic specific call context, for example language, user agent or auth token.
import pytest
from nameko.contextdata import Language
from nameko.rpc import rpc
from nameko.testing.services import entrypoint_hook
class HelloService:
""" Service under test
"""
name = "hello_service"
language = Language()
@rpc
def hello(self, name):
greeting = "Hello"
if self.language == "fr":
greeting = "Bonjour"
elif self.language == "de":
greeting = "Gutentag"
return "{}, {}!".format(greeting, name)
@pytest.mark.parametrize("language, greeting", [
("en", "Hello"),
("fr", "Bonjour"),
("de", "Gutentag"),
])
def test_hello_languages(language, greeting, container_factory, rabbit_config):
container = container_factory(HelloService, rabbit_config)
container.start()
context_data = {'language': language}
with entrypoint_hook(container, 'hello', context_data) as hook:
assert hook("Matt") == "{}, Matt!".format(greeting)
entrypoint_waiter¶
The entrypoint waiter is a context manager that does not exit until a named entrypoint has fired and completed. This is useful when testing integration points between services that are asynchronous, for example receiving an event:
from nameko.events import event_handler
from nameko.standalone.events import event_dispatcher
from nameko.testing.services import entrypoint_waiter
class ServiceB:
""" Event listening service.
"""
name = "service_b"
@event_handler("service_a", "event_type")
def handle_event(self, payload):
print("service b received", payload)
def test_event_interface(container_factory, rabbit_config):
container = container_factory(ServiceB, rabbit_config)
container.start()
dispatch = event_dispatcher(rabbit_config)
# prints "service b received payload" before "exited"
with entrypoint_waiter(container, 'handle_event'):
dispatch("service_a", "event_type", "payload")
print("exited")
Note that the context manager waits not only for the entrypoint method to complete, but also for any dependency teardown. Dependency-based loggers such as (TODO: link to bundled logger) for example would have also completed.
Using pytest¶
Nameko’s test suite uses pytest, and makes some useful configuration and fixtures available for your own tests if you choose to use pytest.
They are contained in nameko.testing.pytest
. This module is automatically registered as a pytest plugin by setuptools if you have pytest installed.
Writing Extensions¶
Structure¶
Extensions should subclass nameko.extensions.Extension
. This base class provides the basic structure for an extension, in particular the following methods which can be overridden to add functionality:
-
Extension.
setup
()¶ Called on bound Extensions before the container starts.
Extensions should do any required initialisation here.
-
Extension.
start
()¶ Called on bound Extensions when the container has successfully started.
This is only called after all other Extensions have successfully returned from
Extension.setup()
. If the Extension reacts to external events, it should now start acting upon them.
-
Extension.
stop
()¶ Called when the service container begins to shut down.
Extensions should do any graceful shutdown here.
Writing Dependency Providers¶
Almost every Nameko application will need to define its own dependencies – maybe to interface with a database for which there is no community extension or to communicate with a specific web service.
Dependency providers should subclass nameko.extensions.DependencyProvider
and implement a get_dependency()
method that returns the object to be injected into service workers.
The recommended pattern is to inject the minimum required interface for the dependency. This reduces the test surface and makes it easier to exercise service code under test.
Dependency providers may also hook into the worker lifecycle. The following three methods are called on all dependency providers for every worker:
-
DependencyProvider.
worker_setup
(worker_ctx)¶ Called before a service worker executes a task.
Dependencies should do any pre-processing here, raising exceptions in the event of failure.
Example: ...
Parameters: - worker_ctx : WorkerContext
See
nameko.containers.ServiceContainer.spawn_worker
-
DependencyProvider.
worker_result
(worker_ctx, result=None, exc_info=None)¶ Called with the result of a service worker execution.
Dependencies that need to process the result should do it here. This method is called for all Dependency instances on completion of any worker.
Example: a database session dependency may flush the transaction
Parameters: - worker_ctx : WorkerContext
See
nameko.containers.ServiceContainer.spawn_worker
-
DependencyProvider.
worker_teardown
(worker_ctx)¶ Called after a service worker has executed a task.
Dependencies should do any post-processing here, raising exceptions in the event of failure.
Example: a database session dependency may commit the session
Parameters: - worker_ctx : WorkerContext
See
nameko.containers.ServiceContainer.spawn_worker
Concurrency and Thread-Safety¶
The object returned by get_dependency()
should be thread-safe, because it may be accessed by multiple concurrently running workers.
The worker lifecycle are called in the same thread that executes the service method. This means, for example, that you can define thread-local variables and access them from each method.
Example¶
A simple DependencyProvider
that sends messages to an SQS queue.
from nameko.extensions import DependencyProvider
import boto3
class SqsSend(DependencyProvider):
def __init__(self, url, region="eu-west-1", **kwargs):
self.url = url
self.region = region
super(SqsSend, self).__init__(**kwargs)
def setup(self):
self.client = boto3.client('sqs', region_name=self.region)
def get_dependency(self, worker_ctx):
def send_message(payload):
# assumes boto client is thread-safe for this action, because it
# happens inside the worker threads
self.client.send_message(
QueueUrl=self.url,
MessageBody=payload
)
return send_message
Writing Entrypoints¶
You can implement new Entrypoint extensions if you want to support new transports or mechanisms for initiating service code.
The minimum requirements for an Entrypoint are:
- Inherit from
nameko.extensions.Entrypoint
- Implement the
start()
method to start the entrypoint when the container does. If a background thread is required, it’s recommended to use a thread managed by the service container (see Spawning Background Threads) - Call
spawn_worker()
on the bound container when appropriate.
Example¶
A simple Entrypoint
that receives messages from an SQS queue.
from nameko.extensions import Entrypoint
from functools import partial
import boto3
class SqsReceive(Entrypoint):
def __init__(self, url, region="eu-west-1", **kwargs):
self.url = url
self.region = region
super(SqsReceive, self).__init__(**kwargs)
def setup(self):
self.client = boto3.client('sqs', region_name=self.region)
def start(self):
self.container.spawn_managed_thread(
self.run, identifier="SqsReceiver.run"
)
def run(self):
while True:
response = self.client.receive_message(
QueueUrl=self.url,
WaitTimeSeconds=5,
)
messages = response.get('Messages', ())
for message in messages:
self.handle_message(message)
def handle_message(self, message):
handle_result = partial(self.handle_result, message)
args = (message['Body'],)
kwargs = {}
self.container.spawn_worker(
self, args, kwargs, handle_result=handle_result
)
def handle_result(self, message, worker_ctx, result, exc_info):
# assumes boto client is thread-safe for this action, because it
# happens inside the worker threads
self.client.delete_message(
QueueUrl=self.url,
ReceiptHandle=message['ReceiptHandle']
)
return result, exc_info
receive = SqsReceive.decorator
Used in a service:
from .sqs_receive import receive
class SqsService(object):
name = "sqs-service"
@receive('https://sqs.eu-west-1.amazonaws.com/123456789012/nameko-sqs')
def handle_sqs_message(self, body):
""" This method is called by the `receive` entrypoint whenever
a message sent to the given SQS queue.
"""
print(body)
return body
Expected Exceptions¶
The Entrypoint base class constructor will accept a list of classes that should be considered to be “expected” if they are raised by the decorated service method. This can used to differentiate user errors from more fundamental execution errors. For example:
class Service:
name = "service"
auth = Auth()
@rpc(expected_exceptions=Unauthorized)
def update(self, data):
if not self.auth.has_role("admin"):
raise Unauthorized()
# perform update
raise TypeError("Whoops, genuine error.")
The list of expected exceptions are saved to the Entrypoint instance so they can later be inspected, for example by other extensions that process exceptions, as in nameko-sentry.
Sensitive Arguments¶
In the same way as expected exceptions, the Entrypoint constructor allows you to mark certain arguments or parts of arguments as sensitive. For example:
class Service:
name = "service"
auth = Auth()
@rpc(sensitive_arguments="password", expected_exceptions=Unauthenticated)
def login(self, username, password):
# raises Unauthenticated if username/password do not match
return self.auth.authenticate(username, password)
This can to be used in combination with the utility function nameko.utils.get_redacted_args()
, which will return an entrypoint’s call args (similar to inspect.getcallargs()
) but with sensitive elements redacted.
This is useful in Extensions that log or save information about entrypoint invocations, such as nameko-tracer.
For entrypoints that accept sensitive information nested within an otherwise safe argument, you can specify partial redaction. For example:
# by dictionary key
@entrypoint(sensitive_arguments="foo.a")
def method(self, foo):
pass
>>> get_redacted_args(method, foo={'a': 1, 'b': 2})
... {'foo': {'a': '******', 'b': 2}}
# list index
@entrypoint(sensitive_arguments="foo.a[1]")
def method(self, foo):
pass
>>> get_redacted_args(method, foo=[{'a': [1, 2, 3]}])
... {'foo': {'a': [1, '******', 3]}}
Slices and relative list indexes are not supported.
Spawning Background Threads¶
Extensions needing to execute work in a thread may choose to delegate the management of that thread to the service container using spawn_managed_thread()
.
def start(self):
self.container.spawn_managed_thread(
self.run, identifier="SqsReceiver.run"
)
Delegating thread management to the container is advised because:
- Managed threads will always be terminated when the container is stopped or killed.
- Unhandled exceptions in managed threads are caught by the container and will cause it to terminate with an appropriate message, which can prevent hung processes.
More Information¶
About Microservices¶
An approach to designing software as a suite of small services, each running in its own process and communicating with lightweight mechanisms.
—Martin Fowler, Microservices Architecture
Microservices are usually described in contrast to a “monolith” – an application built as a single unit where changes to any part of it require building and deploying the whole thing.
With microservices, functionality is instead split into “services” with well defined boundaries. Each of these services can be developed and deployed individually.
There are many benefits as well as drawbacks to using microservices, eloquently explained in Martin Fowler’s paper. Not all of them always apply, so below we’ll outline some that are relevant to Nameko.
Benefits¶
Small and single-purpose
Breaking a large application into smaller loosely-coupled chunks reduces the cognitive burden of working with it. A developer working on one service isn’t required to understand the internals of the rest of the application; they can rely on a higher-level interface of the other services.
This is harder to achieve in a monolithic application where the boundaries and interfaces between “chunks” are fuzzier.
Explicit published interface
The entrypoints for a Nameko service explicitly declare its published interface. This is the boundary between the service and its callers, and thus the point beyond which backwards compatibility must be considered or maintained.
Individually deployable
Unlike a monolith which can only be released all at once, Nameko services can be individually deployed. A change in one service can be made and rolled out without touching any of the others. Long running and highly considered release cycles can be broken into smaller, faster iterations.
Specialization
Decoupled chunks of application are free to use specialized libraries and dependencies. Where a monolith might be forced to choose a one-size-fits-all library, microservices are unshackled from the choices made by the rest of the application.
Drawbacks¶
Overhead
RPC calls are more expensive than in-process method calls. Processes will spend a lot of time waiting on I/O. Nameko mitigates wastage of CPU cycles with concurrency and eventlet, but the latency of each call will be longer than in a monolithic application.
Cross-service transactions
Distributing transactions between multiple processes is difficult to the point of futility. Microservice architectures work around this by changing the APIs they expose (see below) or only offering eventual consistency.
Coarse-grained APIs
The overhead and lack of transactions between service calls encourages coarser APIs. Crossing service boundaries is expensive and non-atomic.
Where in a monolithic application you might write code that makes many calls to achieve a certain goal, a microservices architecture will encourage you to write fewer, heavier calls to be atomic or minimize overhead.
Understanding interdependencies
Splitting an application over multiple separate components introduces the requirement to understand how those components interact. This is hard when the components are in different code bases (and developer head spaces).
In the future we hope to include tools in Nameko that make understanding, documenting and visualizing service interdependencies easier.
Further Notes¶
Microservices can be adopted incrementally. A good approach to building a microservices architecture is to start by pulling appropriate chunks of logic out of a monolithic application and turning them into individual services.
Benefits of Dependency Injection¶
The dependency injection pattern in nameko facilitates a separation of concerns between component parts of a service. There is a natural division between “service code” – application logic tied to the single-purpose of the service – and the rest of the code required for the service to operate.
Say you had a caching service that was responsible for reading and writing from memcached, and included some business-specific invalidation rules. The invalidation rules are clearly application logic, whereas the messy network interface with memcached can be abstracted away into a dependency.
Separating these concerns makes it easier to test service code in isolation. That means you don’t need to have a memcached cluster available when you test your caching service. Furthermore it’s easy to specify mock responses from the memcached cluster to test invalidation edge cases.
Separation also stops test scopes bleeding into one another. Without a clear interface between the caching service and the machinery it uses to communicate with memcached, it would be tempting to cover network-glitch edge cases as part of the caching service test suite. In fact the tests for this scenario should be as part of the test suite for the memcached dependency. This becomes obvious when dependencies are used by more than one service – without a separation you would have to duplicate the network-glitch tests or seem to have holes in your test coverage.
A more subtle benefit manifests in larger teams. Dependencies tend to encapsulate the thorniest and most complex aspects of an application. Whereas service code is stateless and single-threaded, dependencies must deal with concurrency and thread-safety. This can be a helpful division of labour between junior and senior developers.
Dependencies separate common functionality away from bespoke application logic. They can be written once and re-used by many services. Nameko’s community extensions aims to promote sharing even between teams.
Getting in touch¶
If you have questions for the Nameko community or developers, there are a number of ways to get in touch:
Mailing list¶
For help, comments or questions, please use the mailing list on google groups.
Contributing¶
Nameko is developed on GitHub and contributions are welcome.
Use GitHub issues to report bugs and make feature requests.
You’re welcome to fork the repository and raise a pull request with your contributions.
You can install all the development dependencies using:
pip install -e .[dev]
and the requirements for building docs using:
pip install -e .[docs]
Pull requests are automatically built with Travis-CI. Travis will fail the build unless all of the following are true:
- All tests pass
- 100% line coverage by tests
- Documentation builds successfully (including spell-check)
See getting in touch for more guidance on contributing.
License¶
Nameko is licensed under the Apache License, Version 2.0. Please see LICENSE in the project root for details.
Release Notes¶
Here you can see the full list of changes between nameko versions. Versions are in form of headline.major.minor numbers. Backwards-compatible changes increment the minor version number only.
Version 2.12.0¶
Released: 2019-03-18
- Refactor utils so standalone.events does not import eventlet (#580)
- Compatibility with latest dependencies (moto #577, pyyaml and kombu #612)
- Timer now waits for the spawned entrypoint to complete before firing again, as documented (#579, #303)
- Timer is also improved to avoid drift (#614)
- Hide password in logged amqp uri (#582)
- Docs updates (#587, #591, #276, #596)
- Nameko shell changed to not catch exceptions when used in non-TTY mode (#597)
Version 2.11.0¶
Released: 2018-08-09
- Compatibility with kombu 4 and pyamqp 2+, minimum supported kombu version is now 4.2 (#564)
Version 2.9.1¶
Released: 2018-07-20
- SSL connections now supported by all AMQP extensions, configurable using the AMQP_SSL config key. (#524)
- Restore compatibility with eventlet 0.22+ (#556)
- Log unhandled worker exceptions at ERROR level instead of INFO (#547)
Version 2.9.0¶
Released: 2018-05-30
- RPC reply queues are now set to expire rather than auto-delete, and are no longer exclusive, allowing clients reconnect. Fixes #359.
- It’s now possible to accept messages in multiple serialization formats. Adds config-based mechanism for specifying custom serializers. See #535.
- Enhanced environment variable substitution including recursive references. See #515.
Version 2.8.5¶
Released: 2018-03-15
- Workaround for a Kombu bug causing new sockets to sometimes have short timeouts. (#521)
Version 2.8.4¶
Released: 2018-02-18
- Fixes a bug where the container crashed if the connection to RabbitMQ was lost while an AMQP entrypoint was running (#511)
- Correction to WorkerContext.immediate_call_id which actually referred to the id of original call. Adds WorkerContext.origin_call_id to replicate the previous behaviour.
Version 2.8.2¶
Released: 2017-12-11
- Remove the pytest –log-level argument added by the pytest plugin since this conflicts with newer versions of pytest (>= 3.3.0). For older versions this can be restored by installing the pytest-catchlog package.
Version 2.8.1¶
Released: 2017-11-29
- Added the ‘show-config’ command which will print the service configuration to the console, after environment variable substitution.
Version 2.8.0¶
Released: 2017-10-31
- Environment variables substituted into config files are now interpreted as YAML rather than bare values, allowing use of rich data types.
Version 2.7.0¶
Released: 2017-10-07
- Set stopped flag in register_provider() to allow PollingQueueConsumer object reuse [fixes #462]
- Refactor of AMQP message publishing logic into nameko.amqp.publish
- Exposes delivery options and other messaging configuration to AMQP-based DependencyProviders. [addresses #374]
- Class attributes for configuring use_confirms, retry and retry_policy have been deprecated from the Publisher, EventDispatcher, and RPC MethodProxy classes. If you were subclassing these classes to set these options, you should now set them at class instantiation time.
Version 2.6.0¶
Released: 2017-04-30
- Environment variables are now interpreted as native YAML data types rather than just strings
- The WSGI Server now uses an explicit logger so it can be controlled using a logging config
- Drops several backwards-compatibility shims that were marked as being maintained only until this release
Version 2.5.4¶
Released: 2017-04-20
- Don’t block the QueueConsumer thread on the worker pool, which could cause service deadlock and occasional dropped AMQP connections [fixes #428]
- Revert prefetch count change from 2.5.3 and ack messages outside of the QueueConsumer thread. [fixes #417 more robustly]
Version 2.5.3¶
Released: 2017-03-16
- Bump the amqp prefetch_count by 1 to max_workers + 1 to fix throughput issue. [fixes 417]
Version 2.5.2¶
Released 2017-02-28
- Improves teardown speed of the rabbit_config pytest fixture
- Support for providing an alternative reply listener to the standalone RPC proxy
Version 2.5.1¶
Released 2017-01-19
- Adds a DependencyProvider to give services access to the config object
- Internal refactor to make all worker lifecycle steps run in the same thread
Version 2.5.0¶
Released 2016-12-20
- Enables publish confirms by default for all AMQP message publishers
- Refactors common AMQP connection code into nameko.amqp
Version 2.4.4¶
Released 2016-11-28
- Adds AMQP heartbeats to Consumer connections
- Handles an uncaught exception caused by a fast-disconnecting client under Python 3 [fixes #367]
Version 2.4.3¶
Released 2016-11-16
- Pins kombu back to a compatible release (<4) [fixes #378]
- Fixes compatibility with latest bpython and ipython shells [fixes #355 and #375]
- Fixes socket cleanup bug in websocket hub [fixes #367]
Version 2.4.2¶
Released 2016-10-10
Added support for environment variables in YAML config files
Enhanced
entrypoint_waiter()
. The new implementation is backwards compatible but additionally:- Gives you access to the result returned (or exception raised)
- Adds the ability to wait for a specific result
- Doesn’t fire until the worker is completely torn down
Version 2.4.1¶
Released 2016-09-14
- Enhanced :class: ~nameko.web.server.WebServer with get_wsgi_app and get_wsgi_server to allow easy usage of WSGI middleware and modifications of the WSGI server.
- Enhanced
replace_dependencies()
to allow specific replacement values to be provided with named arguments.
Version 2.4.0¶
Released 2016-08-30
- Add dictionary access to
standalone.rpc.ClusterProxy
to allow the proxy to call services whose name is not a legal identifier in python (e.g. name has a-
in it). - Add the ability to specify a custom ServiceContainer class via config key. Deprecate the keyword arguments to ServiceRunner and run_services for the same purpose.
- Deprecate the keyword arguments to run_services, ServiceContainer and ServiceRunner.add_service for specifying a custom WorkerContext class. Custom WorkerContext classes can now only be specified with a custom ServiceContainer class that defines the worker_ctx_cls attribute.
- Remove the context_keys attribute of the WorkerContext, which was previously used to “whitelist” worker context data passed from call to call. It was a feature that leaked from a specific implementation into the main framework, and not useful enough in its own right to continue to be supported.
- Refactor ServiceContainer internals for better separation between “managed” and “worker” threads. Improved logging when threads are killed.
Version 2.3.1¶
Released 2016-05-11
- Deprecate
MethodProxy.async
in favour ofMethodProxy.call_async
in preparation for async becoming a keyword - Add support for loading logging configuration from
config.yaml
Version 2.3.0¶
Released 2016-04-05
- Add support for loading configuration file in
nameko shell
via--config
option - Changed
HttpRequestHandler
to allow override how web exceptions are handled - Enabled reliable delivery on broadcast events when combined with a custom
broadcast_identity
. Reliable delivery now defaults to enabled for all handler types. It must be explicitly turned off with broadcast mode unless you override the default broadcast_identity. - Update bundled pytest fixtures to use a random vhost in RabbitMQ by default
- Now requires eventlet>=0.16.1 because older versions were removed from PyPI
Version 2.2.0¶
Released 2015-10-04
- Add support for alternative serializers in AMQP messages
- Add pytest plugin with common fixtures
- Fix examples in documentation and add tests to prevent future breakage
- Fix bug handling non-ascii characters in exception messages
- Various documentation fixes
Version 2.1.1¶
Released 2015-05-11
- Nameko shell to use bpython or ipython interpreter if available
- Support for marking entrypoint arguments as sensitive (for later redaction)
Version 2.1.0¶
Released 2015-04-13
- Changed default AMQP URI so examples work with an unconfigured RabbitMQ.
- Heuristic messages for AMQP connection errors.
- Added six to requirements.
- Minor documentation fixes.
Version 2.0.0¶
Released 2015-03-31
- python 3 compatibility
- Added HTTP entrypoints and experimental websocket support (contributed by Armin Ronacher)
- Added CLI and console script
- Introduction of nameko “extensions” and nomenclature clarification
- Removal of
DependencyFactory
in favour of prototype pattern - Complete documentation rewrite
- Spun out
nameko.contrib.sqlalchemy
into nameko-sqlalchemy as a community extension. - Spun out
nameko.legacy
package into nameko-nova-compat - Rename the standalone rpc proxy to
ServiceRpcProxy
and add aClusterRpcProxy
, using a single reply queue for communicating with multiple remote services. - Make the standalone event dispatcher more shell-friendly, connecting on demand.
Version 1.14.0¶
Released 2014-12-19
- Remove parallel provider in favour of async RPC
- Update
worker_factory()
to raise if asked to replace a non-existent injection. - Add various
__repr__
methods for better logging - Support for timeouts in the (non-legacy) standalone RPC proxy
- Add helper for manipulating an AMQP URI into a dict
Version 1.13.0¶
Released 2014-12-02
- RPC reply queues now auto-delete.
- Extra protection against badly-behaved dependencies during container kill
- Make legacy
NovaRpcConsumer
more robust against failures in theNovaRpc
provider.
Version 1.12.0¶
Released 2014-11-25
- Add ability to make asynchronous rpc calls using the rpc proxy.
- Add a new nameko context key
user_agent
to support including such info in the rpc header.
Version 1.11.5¶
Released 2014-11-18
- Have the standalone rpc proxy mark its reply queues as auto-delete, to stop them staying around after use.
Version 1.11.4¶
Released 2014-11-10
- Make
RpcConsumer
more robust against failures in theRpc
provider. - Add a new exception
MalformedRequest
that RPC providers can raise if they detect an invalid message. Raise this exception in the defaultRpc
provider ifargs
orkwargs
keys are missing from the message. - Fix issues in queue consumer tests against non-localhost brokers.
- Upgrade to eventlet 0.15.2.
- Include pyrabbit in
requirements.txt
(no longer just for tests). - Catch dying containers in the entrypoint_hook to avoid hangs.
- Add
expected_exceptions
kwarg to the rpc entrypoint to enable different exception handling (in dependencies) for user vs system errors.
Version 1.11.2¶
Released 2014-09-18
- Add a default implementation for
acquire_injection
(returningNone
) for dependency providers that are used for side-effects rather than injecting dependencies.
Version 1.11.1¶
Released 2014-09-15
- New test helper
entrypoint_waiter()
to wait for entrypoints (e.g. event handlers) to complete.
Version 1.11.0¶
Released 2014-09-01
- Raise a specific
RpcTimeout
error in the RPC proxy rather thansocket.timeout
to avoid confusing kombu’sConnection.ensure
- Improve logging helpers
- Use
inspect.getcallargs
instead of shadow lambda for RPC argument checking - Add default retry policies to all publishers
- Stricter handling of connections between tests
- Workarounds for RabbitMQ bugs described at https://groups.google.com/d/topic/rabbitmq-users/lrl0tYd1L38/discussion
Version 1.10.1¶
Released 2014-08-27
- Inspect the service class (instead of an instance) in
worker_factory()
. Works better with descriptors. - Explicitly delete
exc_info
variable when not needed, to help the garbage collector.
Version 1.10.0¶
Released 2014-08-14
- Entrypoint providers’ handle_result is now able to manipulate and modify and
return the
(result, exc_info)
tuple. This enables default post-processing (e.g. serialization, translations) - Added serialization safety to legacy RPC entrypoint.
Version 1.9.0¶
Released 2014-07-15
- No longer relying on eventlet for standalone RPC proxy timeouts.
- Introduced RPC entrypoints compatible with the ‘legacy’ proxy.
Version 1.8.2¶
Released 2014-07-07
- Documentation generator accepts a function listing event classes and adds to output accordingly.
Version 1.8.0¶
Released 2014-06-13
- Now passing
exc_info
tuples instead of bare exceptions toworker_result
andhandle_result
, to enable exception processing in non-worker greenthreads.
Version 1.7.2¶
Released 2014-06-10
_run_worker()
now calls anyhandle_result
method before dependency teardown.- Serialization errors now generate a specific error message rather than bubbling into the container.
- Minor change to
nameko_doc
output.
Version 1.7.1¶
Released 2014-05-20
- Added
language
,auth_token
anduser_id
dependency providers to make context data available to service workers. - Refactored constants into their own module.
- Minor test changes to enable testing on shared rabbit brokers.
Version 1.7.0¶
Released 2014-05-07
spawn_worker()
now throwsContainerBeingKilled
if akill
is in progress, since some providers may already be dead. Providers should catch this and e.g. requeue rpc messages. There is a race condition between completing the kill sequence and remaining entrypoints firing.
Version 1.6.1¶
Released 2014-04-03
- Revert changes to legacy exception serialization to maintain backwards compatibility with old clients.
- Add forwards compatibility for future clients that wish to serialize exceptions into more data
- Promote conftest rabbit manipulations to test helpers
Version 1.6.0¶
Released 2014-03-31
- Rename instance_factory to worker_factory
- Raise
IncorrectSignature
instead ofRemoteError: TypeError
if an RPC method is called with invalid arguments. - Raise
MethodNotFound
instead ofRemoteError: MethodNotFound
if a non-existent RPC method is called. - Let log handlers format warning messages so that aggregators group them correctly.
- Expose the entire dependency provider (rather than just the method name) to the worker context.
Version 1.5.0¶
Released 2014-03-27
Improvements to
kill()
enabling better tracebacks and cleaner teardown:- Using
sys.exc_info
to preserve tracebacks - No longer passing exception into
kill()
, removing race conditions. - No longer requiring
exc
inkill()
- Using
Version 1.4.1¶
Released 2014-03-26
- Adds the
nameko_doc
package, for easing the creation of service-oriented documentation.
Version 1.4.0¶
Released 2014-03-20
- RPC calls to non-existent services (no queues bound to the RPC exchange with the appropriate routing key) now raise an exception instead of hanging indefinitely. Note that calls to existing but non-running services (where the queue exists but has no consumer) behave as before.
Version 1.3.5¶
Released 2014-03-05
- Increased test resilience. Force-closing existing connections on rabbit reset
Version 1.3.4¶
Released 2014-03-05
- Use
MagicMock
for dependency replacement in test utilities - Use
autospec=True
wherever possible when mocking - Merge
ServiceContainers
into a single class
Version 1.3.3¶
Released 2014-02-25
- Bugfixes enabling reconnection to the broker if the connection is temporarily lost.
Version 1.3.2¶
Released 2014-02-13
- Dropping headers with a
None
value because they can’t be serialized by AMQP
Version 1.3.1¶
Released 2014-01-28
- Add
event_handler_cls
kwarg to theevent_handler
entrypoint, for using a custom subclass of theEventHandler
provider
Version 1.3.0¶
Released 2014-01-23
- Standalone RPC proxy interface changed to class with contextmanager interface
and manual
start()
andstop()
methods.
\ Sort by:\ best rated\ newest\ oldest\
\\
Add a comment\ (markup):
\``code``
, \ code blocks:::
and an indented block after blank line