Goal
Kuryr uses it's own way of running controller loops. This makes it impossible to implement some good patterns known from the Kubernetes controllers using controller-runtime.

Problem
There are two main elements missing in Kuryr.

• Ability to specify when to retry an event and using retrying pretty extensively.
• Making sure no failure is fatal for Kuryr.
• Maybe with exception of VLAN IDs exhaustion.

Why is this important
cluster-api-provider-openstack and openstack-cloud-provider present many great patterns on how to deal with reliable creation and management of OpenStack resources. Due to Kuryr limitations we cannot implement these patterns. Improvements in that would allow us to increase reliability. Unfortunately it's not magic bullet, some of the patterns are still subject to issues induced by high scale of requests made by Kuryr.

Estimate (XS, S, M, L, XL, XXL): M