Bug 1754261

Created attachment 1617755 [details]
DataVolume spec

Description of problem:
VMI, with DataVolume defined for its disk, starts successfully and is in "Running" state, but its virtctl console does not function.


Version-Release number of selected component (if applicable):
OCP: 4.2.0-0.nightly-2019-09-21-183303
Latest CNV v2.1.0 (due to 2019-Sep-22)

How reproducible:
Always

Steps to Reproduce:
1. Create DataVolume using the attached dv.yaml
$ oc create -f dv.yaml

2. Verify the DataVolume was created successfully:
 $ oc get dv
NAME        AGE
dv-cirros   23m

3. Verify PVC was created successfully, and that it is in bound state:
$ oc get pvc
NAME        STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS      AGE
dv-cirros   Bound    pvc-93b30aee-dd1e-11e9-af4b-fa163e474cdd   500Mi      RWO            rook-ceph-block   23m

4. Create a Vm that deploys this DataVolume. You can use the attached vm-dv-cirros.yaml
$ oc create -f vm-dv-cirros.yaml

5. Start the VMI:
$ virtctl start vm-cirros

6. Verify the VMI is in "Running" state (you might want to wait ~1 minute for that):
$ oc get vmi
NAME         AGE   PHASE     IP            NODENAME
vm-cirros    24m   Running   10.129.0.41   host-172-16-0-15

7. Open a console to the VMI:
$ virtctl console vm-cirros

Actual results:
The console is not responsive - there's no progress of the VM start-up, no login and no shell.

Expected results:
Interactive console, with login and shell CLI.


Additional info:
1. Same thing happens when using a VM with a non-block (file-system) DV (same DV spec, without volumeMode defined).
2. There's no problem when using the standard vm-cirros spec - the one which is in https://github.com/kubevirt/kubevirt/tree/master/examples, which uses a containerDisk image from the registry.