Improve control plane performance of the Port deletion flows
Registered by
Antoni Segura Puimedon
We need to speed up a lot the performance of the time between the Pod being marked deleted in the Kubernetes API and CNI returning with the port already deleted.
This implies two things:
- Return used ports as quickly and cheaply as possible to the pool if pools are in use.
- Reduce the work and calls done in the CNI side (Possibly caching the data about running vifs in the worker node fs).
Blueprint information
- Status:
- Started
- Approver:
- Antoni Segura Puimedon
- Priority:
- Essential
- Drafter:
- Antoni Segura Puimedon
- Direction:
- Approved
- Assignee:
- Antoni Segura Puimedon
- Definition:
- Approved
- Series goal:
- Accepted for queens
- Implementation:
- Started
- Milestone target:
- queens-2
- Started by
- Antoni Segura Puimedon
- Completed by
Related branches
Related bugs
Sprints
Whiteboard
Addressed by: https:/
Optimize return_to_pool actions at Pool drivers
Gerrit topic: https:/
Addressed by: https:/
Optimize return_to_pool actions at Pool drivers
(?)