-
Bug
-
Resolution: Done
-
Critical
-
None
-
4.14
-
No
-
SDN Sprint 241
-
1
-
Rejected
-
False
-
kube-apiserver cpu usage increased to 34% in OVN IC compared to openshiftSDN. This is observed while running node-density-cni with 80 pods per node configuration on a 120 node environment. Below are the kube-apiserver cpu usage numbers in all the 3 environments -
SDN | Avg CPU % | Max CPU % | |||||
---|---|---|---|---|---|---|---|
master-0 | master-1 | master-2 | Total | master-0 | master-1 | master-2 | |
OpenshiftSDN | 178 | 95.4 | 67 | 340.4 | 284 | 151 | 97.1 |
OVN | 146 | 92.5 | 50.7 | 289.2 | 235 | 129 | 93.8 |
OVN-IC | 195 | 147 | 114 | 456 | 280 | 230 | 169 |
Avg kube-apiserver CPU usage 34% increase in IC compared to OpenshiftSDN
Avg kube-apiserver CPU usage 57% increase in IC compared to OVN
Max kube-apiserver CPU usage is also very high compared to both OpenshiftSDN and OVN.
kube-apiserver CPU & memory usage grafana panels -
Openshiftsdn https://grafana.rdu2.scalelab.redhat.com:3000/dashboard/snapshot/Tm54HMwv1m7VOD5T739xygzdqZ6sU3sS
OVN https://grafana.rdu2.scalelab.redhat.com:3000/dashboard/snapshot/m44tpS4xo2b5NQd4MIctTX0Bx3uDb1kN
OVN-IC https://grafana.rdu2.scalelab.redhat.com:3000/dashboard/snapshot/G6CHY96rqdwlHOiwAty1El1fNuX5c2Zc
OCP version 4.14.0-0.nightly-2023-07-30-234232
I can provide the environment when the engineer wants to debug.
- is related to
-
API-1648 ocp 4.14 kube-apiserver dramatic increase in rate of terminated node watchers
- Closed
-
OCPBUGS-18097 OCP 4.14 increased rate of patch nodes requests from node SAs
- Closed
- links to