Description of problem:

Seemingly randomly, sometimes a container push results in two repositories being created, which fails the push with the "Its Container Repository Name (%{container_name}) conflicts with an existing repository." error.

This is due to a race condition. Since APIs are called simultaneously, I was able to catch two puma workers both reaching https://github.com/katello/katello/blob/master/app/controllers/katello/api/registry/registry_proxies_controller.rb#L346 at the same time.

The get_root_repo_from_product check returns true for both workers, so two repositories get created. We need to figure out how to synchronize the simultaneous requests.

How reproducible:

Sometimes, it's a race condition.

Is this issue a regression from an earlier version:

No.

Steps to Reproduce:

1. Stick a debugger before the `get_root_repo_from_product` line in `create_container_repo_if_needed`.

2. Push Quay's prometheus/busybox image

3. In the debugger, see that two separate workers both reach the same line of code. This is an indication that there's a race condition – it's possible that the repository will not be created yet for either.

Actual behavior:

An error about the container repo name conflicting is thrown.

Expected behavior:
The container image is pushed without error.

Business Impact / Additional info:

QE Tracker for https://issues.redhat.com/browse/SAT-27512