Retrospective on SSH Receive Pack Acceptance Testing Failure

In #525 (closed) we spent time to ensure that we were ready to test the *RecievePack endpoints.

However, in our initial testing of #368 (closed), with gitaly_ssh_receive_pack set to 5%, failures on /api/v4/internal/allowed led to us halting the test.

I'd like to devote some time to looking at several things related to this:

• Why, exactly, is /api/v4/internal/allowed failing?

Once we know this, lets look at what we can do to improve our testing process

• What can we do to improve our process and catch these problems earlier?

• What can we improve in our testing to ensure that we catch problems like this earlier?

  1. How can we improve testing in a nightly build environment?
  2. How can we improve staging?
  3. Could we improve out docker-compose setup?