-
Bug
-
Resolution: Done-Errata
-
Major
-
None
-
4.12, 4.11
-
Important
-
No
-
CNF Network Sprint 247
-
1
-
False
-
This is a clone of issue OCPBUGS-23059. The following is the description of the original issue:
—
This is a clone of issue OCPBUGS-19440. The following is the description of the original issue:
—
This is a clone of issue OCPBUGS-8266. The following is the description of the original issue:
—
Description of problem:
Sriov networknodepolicies and networks are first created in the cluster. While running a series of cluster-density tests, target networkNamespaces are created under which pods are added to the previously created sriov networks. It takes abnormally long for the network-attachments to get created after creation of target namespace specifically when cache entries don't exist:
See several instances of these entries in the network-resource-injector pod logs:
I0302 16:48:16.576685 1 webhook.go:746] Received mutation request. Features status: HugePageInject: true / HonorExistingResources: false / EnableResourceNames: true I0302 16:48:16.578497 1 webhook.go:763] AdmissionReview request received for pod boatload-1/ I0302 16:48:16.578637 1 webhook.go:719] search v1.multus-cni.io/default-network in original pod annotations I0302 16:48:16.578821 1 webhook.go:726] search v1.multus-cni.io/default-network in user-defined injections I0302 16:48:16.578909 1 webhook.go:740] v1.multus-cni.io/default-network is not found in either pod annotations or user-defined injections I0302 16:48:16.578988 1 webhook.go:719] search k8s.v1.cni.cncf.io/networks in original pod annotations I0302 16:48:16.579077 1 webhook.go:722] k8s.v1.cni.cncf.io/networks is defined in original pod annotations I0302 16:48:16.579213 1 webhook.go:266] 'network-boatload-1' is not in JSON format: invalid character 'e' in literal null (expecting 'u')... trying to parse as comma separated network selections list I0302 16:48:16.579397 1 webhook.go:370] cache entry not found, retrieving network attachment definition 'boatload-1/network-boatload-1' from api server E0302 16:48:16.586266 1 webhook.go:356] could not get Network Attachment Definition boatload-1/network-boatload-1: the server could not find the requested resource E0302 16:48:16.586583 1 webhook.go:375] could not find network attachment definition 'boatload-1/network-boatload-1': could not get Network Attachment Definition boatload-1/network-boatload-1: the server could not find the requested resource I0302 16:48:16.586982 1 webhook.go:422] sending response to the Kubernetes API server
When the network attachment is finally created/found (~ 5 min later):
I0302 16:53:44.813028 1 webhook.go:746] Received mutation request. Features status: HugePageInject: true / HonorExistingResources: false / EnableResourceNames: true I0302 16:53:44.815881 1 webhook.go:763] AdmissionReview request received for pod boatload-1/ I0302 16:53:44.816906 1 webhook.go:719] search v1.multus-cni.io/default-network in original pod annotations I0302 16:53:44.816921 1 webhook.go:726] search v1.multus-cni.io/default-network in user-defined injections I0302 16:53:44.816928 1 webhook.go:740] v1.multus-cni.io/default-network is not found in either pod annotations or user-defined injections I0302 16:53:44.816935 1 webhook.go:719] search k8s.v1.cni.cncf.io/networks in original pod annotations I0302 16:53:44.816942 1 webhook.go:722] k8s.v1.cni.cncf.io/networks is defined in original pod annotations I0302 16:53:44.816956 1 webhook.go:266] 'network-boatload-1' is not in JSON format: invalid character 'e' in literal null (expecting 'u')... trying to parse as comma separated network selections list I0302 16:53:44.817046 1 webhook.go:380] network attachment definition 'boatload-1/network-boatload-1' found I0302 16:53:44.817084 1 webhook.go:387] resource 'openshift.io/net_resource_boatload_1' needs to be requested for network 'boatload-1/network-boatload-1' I0302 16:53:44.817098 1 webhook.go:823] pod boatload-1/ has resource requests: map[openshift.io/net_resource_boatload_1:1] and node selectors: map[] I0302 16:53:44.817140 1 webhook.go:905] patch after all mutations: [{add /spec/containers/0/env/- {CONTAINER_NAME boatload-1 nil}} {add /spec/containers/0/volumeMounts/- {podnetinfo true /etc/podnetinfo <nil> }} {add /spec/volumes/ - {podnetinfo {nil nil nil nil nil nil nil nil nil nil nil nil nil nil nil &DownwardAPIVolumeSource{Items:[]DownwardAPIVolumeFile{DownwardAPIVolumeFile{Path:labels,FieldRef:&ObjectFieldSelector{APIVersion:,FieldPath:metadata.labels,},Reso urceFieldRef:nil,Mode:nil,},DownwardAPIVolumeFile{Path:annotations,FieldRef:&ObjectFieldSelector{APIVersion:,FieldPath:metadata.annotations,},ResourceFieldRef:nil,Mode:nil,},DownwardAPIVolumeFile{Path:hugepages_1G_request_boatload-1,Field Ref:nil,ResourceFieldRef:&ResourceFieldSelector{ContainerName:boatload-1,Resource:requests.hugepages-1Gi,Divisor:{{1048576 0} {<nil>} BinarySI},},Mode:nil,},DownwardAPIVolumeFile{Path:hugepages_1G_limit_boatload-1,FieldRef:nil,ResourceFi eldRef:&ResourceFieldSelector{ContainerName:boatload-1,Resource:limits.hugepages-1Gi,Divisor:{{1048576 0} {<nil>} BinarySI},},Mode:nil,},},DefaultMode:nil,} nil nil nil nil nil nil nil nil nil nil nil nil}}} {add /spec/nodeSelector map[j etlag:true node-role.kubernetes.io/master:]}] for pod boatload-1/ I0302 16:53:44.817270 1 webhook.go:422] sending response to the Kubernetes API server
Also notice this in the multus-admission-controller pod logs:
2023-03-02T16:53:04.610972543Z I0302 16:53:04.610898 1 webhook.go:117] validating network config spec: \{ "cniVersion":"0.3.1", "name":"network-boatload-1","type":"sriov","vlan":100,"spoofchk":"on","trust":"off","vlanQoS":0,"ipam":{} }
2023-03-02T16:53:04.611200606Z I0302 16:53:04.611170 1 webhook.go:141] spec is not a valid network config list: error parsing configuration list: no 'plugins' key - trying to parse into standalone config
2023-03-02T16:53:04.611290278Z I0302 16:53:04.611265 1 webhook.go:154] AdmissionReview request allowed: Network Attachment Definition '\{"cniVersion":"0.3.1","ipam":{},"name":"network-boatload-1","spoofchk":"on","trust":"off","type":"sriov","vlan":100,"vlanQoS":0}' is valid
Version-Release number of selected component (if applicable):
4.11.25
How reproducible:
Steps to Reproduce:
1. Create sriovNetworkNodePolicy CR
2. Create sriovNetwork CR
3. Wait a few hours and then create target namespaces
4. Add a pod in the target namespace to the Sriov network
5. Watch the status of network-attachment-definitions
Actual results:
network-attachment-definitions get created after a significant delay after creation of target namespaces (when there is no cached entry)
Expected results:
network-attachment-definitions should get created immediately after creation of target namespaces
Additional info:
sriovnetwork, sriovnetworknodepolicy CRs and namespace creation script used along with same pod specs attached, must-gather attached
- clones
-
-
- Closed
-
- is blocked by
-
-
- Closed
-
- links to
-
RHBA-2024:0197
OpenShift Container Platform 4.12.z extras update