Skip to content

Latest commit

 

History

History
225 lines (165 loc) · 9.99 KB

README.md

File metadata and controls

225 lines (165 loc) · 9.99 KB

Docker Virtual Switch (DVS) Tests

Introduction

The Docker Virtual Switch tests allow developers to validate the control plane behavior of new SWSS features without needing an actual network device or switching ASIC.

The DVS tests work by publishing configuration updates to redis (typically Config DB or App DB) and checking that the state of the system is correctly updated by SWSS (typically by checking ASIC DB).

SWSS, Redis, and all the other required components run inside a virtual switch Docker container, meaning that these test cases can be run on any Linux machine - no special hardware required!

Contents

Setting up your test environment

  1. In order to set up your test environment you will need:

    • A machine running Ubuntu 18.04 or 20.04
    • A generic Linux kernel to install team
    • python3

    You can check these dependencies with the following:

    cat /etc/os-release | grep ".*18.04\|20.04.*"
    uname -r | grep generic
    python3 --version
    
  2. Install Docker CE. Be sure to follow the post-install instructions so that you don't need sudo privileges to run docker commands.

  3. Install the external dependencies needed to run the tests.

    sudo modprobe team
    sudo apt install python3-pip net-tools bridge-utils ethtool vlan libnl-nf-3-200 libnl-cli-3-200
    sudo pip3 install docker pytest flaky redis distro dataclasses fstring
    

    If you are running Ubuntu 18.04 you will need to install this package:

    sudo apt install libhiredis0.13
    

    Dash testcases aren't supported in Ubuntu 18.04

    If you are running Ubuntu 20.04 you will need to install this package:

    sudo apt install libhiredis0.14
    

    If you want to run DASH testcases, please download and install the latest ubuntu20.04 dependencies of DASH from Azp.

  4. Install swsscommon.

    sudo dpkg -i libswsscommon_1.0.0_amd64.deb python3-swsscommon_1.0.0_amd64.deb
    

    You can get these two packages by:

  5. Load the docker-sonic-vs.gz file into docker. You can get the image by:

    Once you have the file, you can load it into docker by running docker load < docker-sonic-vs.gz.

Running the tests

cd sonic-swss/tests
sudo pytest

Setting up a persistent testbed

For those developing new features for SWSS or the DVS framework, you might find it helpful to setup a persistent DVS container that you can inspect and make modifications to (e.g. using dpkg -i to install a new version of SWSS to test a new feature).

  1. Download create_vnet.sh.

  2. Setup a virtual server and network. Note: It is highly recommended you include the -n 32 option or you may run into problems running the tests later.

    docker run --privileged -id --name sw debian bash
    sudo ./create_vnet.sh -n 32 sw
    
  3. Start the DVS container.

    docker run --privileged -v /var/run/redis-vs/sw:/var/run/redis --network container:sw -d --name vs docker-sonic-vs
    
  4. You can specify your persistent DVS container when running the tests as follows.

    sudo pytest --dvsname=vs
    

    By default if the DVS has <32 ports (needed by testbed), then test will be aborted. To overcome that the --forcedvs option can be used. This automatically adds the proper number of ports for the tests to run, and cleans up the extra ports after the test is complete.

    sudo pytest --dvsname=vs --forcedvs
    
  5. Additionally, if you need to simulate a specific hardware platform (e.g. Broadcom or Mellanox), you can add this environment variable for hardware SKU when starting the DVS container. Note that this is not a precise 1-to-1 model, and dataplane behavior is not simulated by the DVS.

    -e "HWSKU=Mellanox-SN2700"
    

Other useful test parameters

  • You can specify a maximum amount of cores for the DVS to use (we recommend 2):

    sudo pytest --max_cpu 2
    

    For a persistent DVS:

    docker run --privileged -v /var/run/redis-vs/sw:/var/run/redis --network container:sw -d --name vs --cpus 2 docker-sonic-vs
    

    For specific details about the performance impact of this, see the Docker docs.

  • You can see the output of all test cases that have been run by adding the verbose flag:

    sudo pytest -v
    

    This works for persistent DVS containers as well.

  • You can specify a specific test file, class, or even individual test case when you run the tests:

    sudo pytest <test_file_name>::<test_class_name>::<test_name>
    

    This works for persistent DVS containers as well.

  • You can specify a specific image:tag to use when running the tests without a persistent DVS container:

    sudo pytest --imgname=docker-sonic-vs:my-changes.333
    
  • You can also preserve a non-persistent DVS container for debugging purposes:

    sudo pytest --keeptb
    

    Which should give you something like this in docker ps:

    CONTAINER ID        IMAGE                                                 COMMAND                  CREATED             STATUS              PORTS               NAMES
    10bb406e7475        docker-sonic-vs:sonic-swss-build.1529                 "/usr/bin/supervisord"   3 hours ago         Up 3 hours                              ecstatic_swartz
    edb35e9aa10b        debian:jessie                                         "bash"                   3 hours ago         Up 3 hours                              elegant_edison
    
  • You can automatically retry failed test cases once:

    sudo pytest --force-flaky
    

Known Issues

  • The test run might abort before any cases are run:

    daall@baker:~/sonic-swss/tests$ sudo pytest test_acl.py 
    ============================= test session starts ==============================
    platform linux -- Python 3.6.9, pytest-4.6.9, py-1.9.0, pluggy-0.13.1
    rootdir: /home/daall/sonic-swss/tests
    plugins: flaky-3.7.0
    collected 25 items                                                             
    
    test_acl.py Aborted
    

    When run with the -sv flags we get some more information:

    daall@baker:~/sonic-swss/tests$ sudo pytest -sv test_acl.py 
    ============================= test session starts ==============================
    platform linux -- Python 3.6.9, pytest-4.6.9, py-1.9.0, pluggy-0.13.1 -- /usr/bin/python3
    cachedir: .pytest_cache
    rootdir: /home/daall/sonic-swss/tests
    plugins: flaky-3.7.0
    collected 25 items                                                             
    
    test_acl.py::TestAcl::test_AclTableCreation terminate called after throwing an instance of 'std::runtime_error'
    what():  Sonic database config file doesn't exist at /var/run/redis/sonic-db/database_config.json
    Aborted
    

    This indicates that something went wrong with the libswsscommon installation. The following should mitigate the issue:

    dpkg -r libswsscommon python3-swsscommon
    dpkg --purge libswsscommon python3-swsscommon
    rm -rf /usr/lib/python3/dist-packages/swsscommon/
    dpkg -i libswsscommon_1.0.0_amd64.deb python3-swsscommon_1.0.0_amd64.deb
    
  • The tests might report a syntax error:

    ImportError while loading conftest '/sonic/src/sonic-swss/tests/conftest.py'.
    File "/sonic/src/sonic-swss/tests/conftest.py", line 56
        def __init__(self, db_id: str, connector: str):
    

    This indicates that pytest is using python2 rather than python3. You need to install and use pytest for python3.

  • You may encounter the following error message:

    ERROR: Error response from daemon: client is newer than server (client API version: x.xx, server API version: x.xx)
    

    You can mitigate this by upgrading to a newer version of Docker CE or editing the DEFAULT_DOCKER_API_VERSION in /usr/local/lib/python3/dist-packages/docker/constants.py, or by upgrading to a newer version of Docker CE. See relevant GitHub discussion.

  • Currently when pytest is run using --force-flaky and the last test case fails, then pytest tears down the module before retrying the failed test case and invokes the module setup again to run the failed test case. This is a known issue with flaky and pytest.

    Because of this issue, all the logs are lost except for the last test case as modules are torn down and set up again. The workaround for this is to include a dummy test case that always passes at the end of all test files/modules.