Description of problem: When trying to run the kubevirt apb in an offline mode using the registry_url apb parameter is not used when templating the CDI deployments causing the pods fail to start. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: --- apiVersion: servicecatalog.k8s.io/v1beta1 kind: ServiceInstance metadata: name: kubevirt namespace: kube-system spec: clusterServiceClassExternalName: localregistry-virtualization clusterServicePlanExternalName: default parameters: admin_user: admin admin_password: ... registry_url: docker-registry.default.svc:5000 registry_namespace: openshift version: 0.9.3 docker_tag: v1.3.0 Actual results: Expected results: Additional info: Is this related? https://bugzilla.redhat.com/show_bug.cgi?id=1644689
What registry do the cdi images pull from?
(In reply to Ryan Hallisey from comment #1) > What registry do the cdi images pull from? The openshift registry. In the process of performing an offline install for the lab we are providing we pull from brew then push into the openshift registry. I don't have the cluster that I tested this on but when I looked at the deployments the none of the image parameters had the `registry_url` provided. The other components had no issues.
(In reply to Joseph Callen from comment #2) > (In reply to Ryan Hallisey from comment #1) > > What registry do the cdi images pull from? > > The openshift registry. In the process of performing an offline install for > the lab we are providing we pull from brew then push into the openshift > registry. I don't have the cluster that I tested this on but when I looked > at the deployments the none of the image parameters had the `registry_url` To be specific the CDI deployments > provided. The other components had no issues.
If you do get the environment back up can you paste the image that the cdi deployment attempted to pull from? If I understand what you saying, all the working components images should have as a registry 'docker-registry.default.svc:5000', but cdi deployments do not. I'll see if there are any clues in the code as to what this is.
Rebuilt...I can get you access if you want it. CDI Deployments > oc get deploy -l cdi.kubevirt.io --template '{{ range .items }} {{ range .spec.template.spec.containers }} {{ .image }}{{end}}{{end}}' cnv-tech-preview/virt-cdi-apiserver:v1.3.0 cnv-tech-preview/virt-cdi-controller:v1.3.0 cnv-tech-preview/virt-cdi-uploadproxy:v1.3.0 ^ missing registry_url kubevirt deployments > oc get deploy --template '{{ range .items }} {{ range .spec.template.spec.containers }} {{ .image }}{{end}}{{end}}' docker-registry.default.svc:5000/cnv-tech-preview/virt-api:v1.3.0 docker-registry.default.svc:5000/cnv-tech-preview/virt-controller:v1.3.0 DaemonSets > oc get ds --template '{{ range .items }} {{ range.spec.template.spec.containers }} {{ .image }}{{end}}{{end}}' docker-registry.default.svc:5000/cnv-tech-preview/multus-cni:v1.3.0 docker-registry.default.svc:5000/cnv-tech-preview/ovs-cni-plugin:v1.3.0 docker-registry.default.svc:5000/openshift3/node:latest ^ this is a bug too. This _should_ be ose-node:v3.11 docker-registry.default.svc:5000/cnv-tech-preview/virt-handler:v1.3.0
Thanks. -docker_prefix: "{{ cdi_repo_tag | default('cnv-tech-preview') }}" +docker_prefix: "{{ registry_url }}/{{ registry_namespace }}"
cnv-tech-preview is the correct downstream repo so this bug isn't breaking. Will need acks from PM and QE to build.
we have decided to close it for now, because the issue comes from kubevirt-ansible, and we probably wont hit it with the operators