Loading
Reduce memory requests tests further
The integration-tests still sometimes fail due to no memory for nova-compute. So reduce the request of mainly db and MQ but also some services and API, so that nova-compute can get scheduled on the nodes.
Edit: Additionally fix insufficient memory for neutron-api and octavia cd pods. This change introduces 3 additional compute nodes to the integrationtest and removes single worker node. A priorityClass for octavia cds is introduced to evict pods from a fully utilised node so these pods can be scheduled after a cold migration. (If you wonder why the memory resources use such weird values, that is because most were set with a script using the memory metric from another integrationtest run.)
Edited by Theresa Schüttig