Cache team_size on Team model.

[cahrens]

@dan-f WIP PR.

[cahrens]

@dan-f How will this field get populated for teams with existing memberships?

[cahrens]

Note that this will need a rebase-- my new PR is https://github.com/edx/edx-platform/pull/9503. And migration number 5 changed, so you will likely need to back up and redo migrations.

[peter-fogg]

@cahrens I've removed the is_active field from teams, and added team_size with the appropriate indices.

[cahrens]

How can we be sure that memberships will never be created in this way (this would not update the count, correct)?

[dianakhuang]

This way should update the count, I believe, because the count gets updated on the save of the team membership.

[peter-fogg]

Correct. I changed these lines to call add_user just because it makes sense to test membership creation in the way that the actual application code does it.

[cahrens]

Nice work, @peter-fogg. I just have that one question (wondering if we should handle the addition in the create method of the membership itself). Otherwise 👍, and this is certainly in good enough shape to test with tomorrow.

Please let the devops person who deploys the Ami to the load testing environment know that migrations need to be done.

[peter-fogg]

@cahrens I had originally thought that we couldn't prevent someone from creating a membership with CourseTeamMembership.objects.create, but it turns out we can check in the save method if there's a primary key set on the model. If so, it already existed in the database, but if not, it's being saved for the first time. This should ensure that team sizes are always correct. I've also fixed up some quality violations and squashed commits.

[peter-fogg]

@dianakhuang Can you give a second review if you get a chance?

[dianakhuang]

Nit: This is an odd change, just because it doesn't really do anything differently?

[peter-fogg]

Ah, I'd been experimenting with some things and forgot to change that back. I'll fix it up.

[dianakhuang]

Looks good to me. 👍

[cahrens]

@peter-fogg Since there is no rush with trying to load test today, I'll wait until this PR merges to my branch before rebasing on master.

[peter-fogg]

The only failing JS tests are the known flaky discussions tests, so I'll go ahead and merge this.