Simple networking testbed

[arjo129]

warning This is experimental code and not likely to be stable for
some time.

This repo provides a simple networking test bed using
docker
and docker-compose. You will need to
have both of these tools set up.
Additionally it assumes a checkout from the "discovery_modes_full.repos"

One can test various scenarios use cases using the script and the
docker-compose.yml. To get started simply run:

./test_scenario.bash <path to your ros2 checkout>

This will build a docker image and run a test scenario where a container
publishes a "Hello" message from a test set up. All containers listen on
this setup and report their results to a results folder. In this
repository. You will see the output of rostopic echo /test_topic from
each container in the results folder.

Signed-off-by: Arjo Chakravarty arjo@openrobotics.org

[mxgrey]

This is looking extremely promising.

One thing to note is that we've been advised to test scenarios where the two endpoints have different settings. E.g. one endpoint's environment is restricted to local discovery while another endpoint's environment is set to global discovery. We've been advised that mixing the settings could result in unintended behaviors because of bugs in how the settings are handled and/or because of unintuitive behaviors in how multicast packets are handled by the kernel.

If I'm understanding the current scripts here correctly, this is only testing cases where both endpoints have the same environment settings. We could probably just do a nested for-loop to get all the combinations. And instead of setting RES_DIR in the .env files we would set it as $outer_scenario_name/$inner_scenario_name/. Then we can have an expected_results/ directory which we can recursively diff against the produced results/ directory to determine if the tests passed.

[arjo129]

So I'm currently working on expanding the matrix exactly as you said and a reporting script. I've not yet pushed those changes. I am working on a reporting script as checking every combination manually will be impossible.

…

On Mon, Jan 30, 2023 at 9:13 PM Grey ***@***.***> wrote: This is looking extremely promising. One thing to note is that we've been advised to test scenarios where the two endpoints have different settings. E.g. one endpoint's environment is restricted to local discovery while another endpoint's environment is set to global discovery. We've been advised that mixing the settings could result in unintended behaviors because of bugs in how the settings are handled and/or because of unintuitive behaviors in how multicast packets are handled by the kernel. If I'm understanding the current scripts here correctly, this is only testing cases where both endpoints have the same environment settings. We could probably just do a nested for-loop to get all the combinations. And instead of setting RES_DIR in the .env files we would set it as $outer_scenario_name/$inner_scenario_name/. Then we can have an expected_results/ directory which we can recursively diff against the produced results/ directory to determine if the tests passed. — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAEEMQC5OAFTLU6H3R72CWTWU65AXANCNFSM6AAAAAAT755H6M> . You are receiving this because you authored the thread.Message ID: ***@***.***>