Create Scheduler for Grouping Tests w/ Variable number of Workers

[ciaconet]

This pull request addresses Issue 1014. It adds a new scheduler that allows a user to add tests to arbitrary groups using the pytest mark xdist_custom. Using the mark a group name and number of workers to use for that group is specified. Groups are ran sequentially but tests in each group are ran in parallel with the given number of workers.

This pull request is a rough prototype. We aren't too familiar with pytest-xdist internals and we don't expect it is merge-able as is. However, we're hoping to open the conversation to see if there are improvements we can make to get this merged into pytest-xdist.

Since we wanted groups of tests to execute sequentially, we did not want to schedule tests from a pending group until after the currently running group completed all tests. We found that in order to obtain the status of the final test scheduled on a worker, we needed to send a shutdown signal to the worker. Because of this, between groups a shutdown signal is sent to each worker, and then workers are initialized again prior to scheduling the next group of tests.

We left the directory xdist-testing-ntop in as an example but will remove that and add proper tests.

Will add a changelog/1014.feature if this PR gains traction.

[dmakhno]

@ciaconet , I wonder does it address #385 too?
For my need a concept of pytest-mp (which was archieved) is more suitabe. And this scheduler seems a good fit - and open new strategies for scheduler.

Just my naive visualization on need:

And your CR, "custom marker", is exactly "bucket", "batch".

[ciaconet]

@ciaconet , I wonder does it address #385 too? For my need a concept of pytest-mp (which was archieved) is more suitabe. And this scheduler seems a good fit - and open new strategies for scheduler.

Just my naive visualization on need:

And your CR, "custom marker", is exactly "bucket", "batch".

@ciaconet , I wonder does it address #385 too? For my need a concept of pytest-mp (which was archieved) is more suitabe. And this scheduler seems a good fit - and open new strategies for scheduler.

Just my naive visualization on need:

And your CR, "custom marker", is exactly "bucket", "batch".

Unfortunately as is this PR doesn't address #385. The custom mark would give you the option to run some tests sequentially, but this would happen in it's own "bucket". With the changes on this PR we can't have some tests run sequentially while simultaneously running other tests in parallel.

[vitaly-krugl]

Regarding

between groups a shutdown signal is sent to each worker, and then workers are initialized again prior to scheduling the next group of tests

This may slow down the test run considerably due to test collection overhead when workers start up. My workplace has about 30 thousand test cases at this time - and growing fast - and each test collection cycle takes multiple minutes in our use case.

Each xdist worker performs test collection and filtering independently when it starts up, combined with any additional processing that the test suite might perform via pytest hooks within the scope of each worker.

So, if you shut down and restart all xdist workers between each group, you would inject non-trivial test collection/processing delays between the execution of each group. If someone has many groups, they would incur many such delays.

Please correct me if I am wrong.

[Heroico]

Regarding

between groups a shutdown signal is sent to each worker, and then workers are initialized again prior to scheduling the next group of tests

This may slow down the test run considerably due to test collection overhead when workers start up. My workplace has about 30 thousand test cases at this time - and growing fast - and each test collection cycle takes multiple minutes in our use case.

Each xdist worker performs test collection and filtering independently when it starts up, combined with any additional processing that the test suite might perform via pytest hooks within the scope of each worker.

So, if you shut down and restart all xdist workers between each group, you would inject non-trivial test collection/processing delays between the execution of each group. If someone has many groups, they would incur many such delays.

Please correct me if I am wrong.

Shutting down workers and restarting when changing groups indeeds incurs in collection/processing again in this implementation.

In our case, we have around 20000 test cases, and our collection takes about 10 seconds whenever we switch from one group to the next.