Comment 10 for bug 1999671

Hi Bas; great sleuthing! I want to say thanks for your detailed investigation into this issue. It's incredibly helpful to understand how CPU contention can cause an OpenStack system to start failing, so this is a really good datapoint! There's not a great deal we can do about neutron-api slow down, except to note that it needs to have sufficient CPU headroom to operate; again very useful information for deployment scenarios and great that we can catch this in testing. Nicely done!