-
Bug
-
Resolution: Won't Do
-
Critical
-
None
-
4.12
-
None
-
Important
-
No
-
SDN Sprint 245
-
1
-
False
-
Description of problem:
A customer reported one of the 120 nodes to not being able to start pod for the error: "failed to configure pod interface: timed out waiting for OVS port binding (ovn-installed)" The ovn-controller is reporting: "poll_loop|INFO|wakeup due to [POLLIN][POLLHUP] ... (98% CPU usage)" But the process consuming more CPU on the node is ovn-controller itself.
Version-Release number of selected component (if applicable):
4.12.15
How reproducible:
not known
Steps to Reproduce:
1. install cluster with OVN 2. configure several networkpolicies and services 3.
Actual results:
ovn-controller is consuming too much CPU impacting OVS
Expected results:
ovn-controller CPU usage should not harm the node
Additional info:
- is related to
-
SDN-4194 Define alerts for high OVS CPU usage
- Closed