nuts 3.0.0a1

Network Unit Testing System Component of NetTowel

NetTowel Network Unit Testing System

Introduction

The NetTowel Network Unit Testing System or "nuts" in short is the testing component of the NetTowel Project. It draws on the concept of unit tests, known from the domain of programming, and applies it to the domain of networking.

One major difference between unit tests in programming and network tests is the definition of what a test actually is. In programming, unit tests normally focus on testing edge cases, since the amount of non-edge cases is not definable. In the network testing domain, tests are less about edge cases, but more about testing existing configurations with pre-defined test cases. Such a single test case might be "can host A reach neighbors X, Y, Z?" on many different devices. This is what nuts tries to achieve: Apply test cases based on your pre-defined network topology to your actual network and have the tests confirm the correct configuration.

The project relies on the pytest framework to setup and execute the tests. Nuts itself is written as a custom pytest plugin. In the background, nornir executes specific network tasks for the actual tests.

Additionally, nuts treats the test definition and the so-called test bundle as separate entities.

The test definition is modelled as a custom pytest.Class, and a predefined set of test definitions can be found in the module base_tests. New test definitions can be added easily by the user of the plugin.

The test bundle is a file that is parsed by pytest. The file provides data on the actual network configuration and describes which test definitions should be collected and executed by pytest. The structure of the test bundle should enable people without in-depth python knowledge to add new test bundles or update existing ones to reflect changes in the network.

Test bundle structure

Currently only yaml files are supported as test bundles, but other sources such as other file formats or database entries can be considered in later versions.

Each test bundle contains the following structure:

---
- test_module: <module that contains the test class> # optional
  test_class: <name of the test class>
  label: <label to uniquely identify the test> # optional 
  test_execution: <additional data used to execute the test> # optional
  test_data: <data used to generate the test cases>
...

test_module: The full path of the python module that contains the test class to be used. This value is optional if the test class is registered in index.py of the pytest-nuts plugin. Note that it can be relevant in which directory pytest is started if local test modules are used.

test_class: The name of the python class which contains the tests that should be executed. Note that currently every test in this class will be executed.

label: Additional identifier that can be used to distinguish between multiple occurrences of the same test class in a test bundle.

test_execution: Data that is exposed as part of the nuts_parameters property (explanation see below). By convention this contains additional information that is passed directly to the nornir task in the background. Therefore the key-value pairs must be consistent with the key-value pairs of the specific nornir task. As an example, the test definition napalm_ping.py calls a nornir task to execute napalm's ping-command. This allows the additional max_drop parameter in test execution, since it is in turn pre-defined by napalm.

test_data: Data that is used to parametrize the tests in the test class which have the pytest.mark.nuts annotation. It is additionally part of the nuts_parameters property.

Examples

Example of a test bundle for TestNetmikoCdpNeighbors which tests that R1 is a CDP Neighbor of both R2 and R3. This example creates three different tests, one for each entry in the test_data list.

---
- test_module: nuts.base_tests.netmiko_cdp_neighbors
  test_class: TestNetmikoCdpNeighbors
  test_data:
    - source: R1
      local_port: GigabitEthernet3
      destination_host: R2
      management_ip: 172.16.12.2
      remote_port: GigabitEthernet2
    - source: R1
      local_port: GigabitEthernet4
      destination_host: R3
      management_ip: 172.16.13.3
      remote_port: GigabitEthernet2
    - source: R2
      local_port: GigabitEthernet2
      destination_host: R1
      management_ip: 172.16.12.1
      remote_port: GigabitEthernet3
...

Installation instructions

NetTowel nuts is currently not published via pip. It has to be cloned and installed manually.

git clone ssh://[e-mail]:45022/ins/nettowel/nettowel-nuts.git
pip install <your_nuts_directory>

Technical Overview

Nuts custom marker

During test collection, the custom pytest marker "nuts" uses the data that has been defined in the test bundle mentioned above. This annotation is a wrapper around the pytest.mark.parametrize annotation and allows the plugin to use the data entries from the test bundle. For each entry in the test_data section of the test bundle, the custom marker generates a single test case. Each entry is a dictionary, but pytest.mark.parametrize expects a list of n-tuples as input. The plugin therefore transforms those dictionary entries from test_data into tuples. This transformation is currently fixed, but more flexibility is very likely to come at a later stage.

The custom nuts marker takes two arguments: The first argument of the annotation determines the required fields. For each entry in test_data these fields are extracted and transformed to a tuple considering the correct order. If any of these fields are not present in an entry of test_data, the corresponding test case will be skipped. The second argument determines optional fields that can also be used in a test case as well - non-present values are passed into the function as None.

Example of a test class with custom marker

@pytest.mark.usefixtures("check_nuts_result")  # see below
class TestNetmikoCdpNeighbors:
    @pytest.mark.nuts("host,remote_host,local_port,management_ip", "management_ip")
    def test_local_port(self, single_result, remote_host, local_port):
        assert single_result.result[remote_host]["local_port"] == local_port

This test run of CDP neighbors checks the local port.

Before each test evaluation, the fixture @pytest.mark.usefixtures("check_nuts_result") checks the result of the network information that has been gathered by nornir in the background: It asserts that that no exception was thrown while doing so.

The required fields are host, remote_host and local_port - they must be present in the custom marker, but also be provided as argument to the test method itself. management_ip is an optional field and not necessary for the test - an entry in test_data is not required to have it.

single_result uses the host field and provides the result that has been processed via the specific context of a test.

Test classes and their context

Each test module implements a context class to provide module-specific functionality to its tests. This context class is a NutsContext or a subclass of it. This guarantees a consistent interface across all tests for test setup and execution. Currently, the predefined test classes use nornir in order to communicate with the network devices, therefore the test classes derive all from a more specific NornirNutsContext, which provides a nornir instance and nornir-specific helpers.

This context class is then integrated into pytest fixtures. It is passed on to another fixture called sincle_result that returns a per-host result of one test.

Development

Nuts uses poetry as a dependency manager. If you have not installed poetry, please read their installation instructions.

Installation requirements

poetry install

Open shell with venv

poetry shell

Thanks

• Matthias Gabriel, who laid the foundations of NUTS.
• Florian Bruhin for invaluable feedback.