search

cockpit-project / cockpit

Cockpit is a web-based graphical interface for servers.
http://www.cockpit-project.org/
GNU Lesser General Public License v2.1
10.95k stars 1.1k forks source link

Image refresh for openshift #9479

Closed cockpituous closed 6 years ago

cockpituous commented 6 years ago

Image refresh for openshift

cockpituous commented 6 years ago

image-refresh in progress on cockpit-tasks-zgqmz. Log: http://fedorapeople.org/groups/cockpit/logs/image-refresh-9479-20180623-142929/

cockpituous commented 6 years ago

Task failed: http://fedorapeople.org/groups/cockpit/logs/image-refresh-9479-20180623-142929/

martinpitt commented 6 years ago 
spec.template.spec.domain.cpu.cores in body must be of type integer: "string"
image-create: setup failed with code 1
martinpitt commented 6 years ago

@mlibra: kubevirt 0.6.0's cluster/vm-template-fedora.yaml has:

          cpu:
            cores: ${CPU_CORES}

which is filled in with

oc process -f vm-template-fedora.yaml -p NAME=fedoravm -p MEMORY=256Mi -p CPU_CORES=1 | oc apply -f -

This seems to result in a string result:

"spec\":{\"domain\":{\"cpu\":{\"cores\":\"1\"}

In oc-process(1) I don't see any data type specification. Should kubevirt get along with a string value?

martinpitt commented 6 years ago

I worked around that with sed -i 's/${CPU_CORES}/1/' ./vm-template-fedora.yaml in kubevirt.setup. But now it fails because virt-launcher-fedoravm-zs8kq stays at status Completed forever, and never goes to running.

[root@f1 ~]# oc logs -c volumeregistryvolume virt-launcher-fedoravm-zs8kq
Error from server (BadRequest): container "volumeregistryvolume" in pod "virt-launcher-fedoravm-zs8kq" is waiting to start: trying and failing to pull image

# oc logs -c compute virt-launcher-fedoravm-zs8kq
level=info timestamp=2018-06-26T16:22:42.476109Z pos=monitor.go:209 component=virt-launcher msg="Still missing PID for qemu-system, Process qemu-system not found in /proc"
level=info timestamp=2018-06-26T16:22:42.476205Z pos=monitor.go:213 component=virt-launcher msg="qemu-system not found after timeout"
level=info timestamp=2018-06-26T16:23:12.479194Z pos=virt-launcher.go:218 component=virt-launcher msg="Timed out waiting for final delete notification."
level=info timestamp=2018-06-26T16:23:12.479285Z pos=virt-launcher.go:309 component=virt-launcher msg=Exiting...

For the former, the journal says

openshift-run[10165]: E0626 13:03:40.272530   10174 pod_workers.go:186] Error syncing pod 277ea2e5-795c-11e8-b0e7-525400123456 ("virt-launcher-fedoravm-zs8kq_kubevirt(277ea2e5-795c-11e8-b0e7-525400123456)"), skipping: failed to "StartContainer" for "volumeregistryvolume" with ImagePullBackOff: "Back-off pulling image \"registry:5000/kubevirt/fedora-cloud-registry-disk-demo:latest\""

and there are tons of

E0626 17:04:44.300566       1 reflector.go:205] github.com/kubevirt/containerized-data-importer/pkg/client/informers/externalversions/factory.go:74: Failed to list *v1alpha1.DataVolume: datavolumes.cdi.io is forbidden: User "system:serviceaccount:golden-images:cdi-sa" cannot list datavolumes.cdi.io at the cluster scope: User "system:serviceaccount:golden-images:cdi-sa" cannot list all datavolumes.cdi.io in the cluster