-
Bug
-
Resolution: Unresolved
-
Undefined
-
None
-
None
-
0.42
-
False
-
-
False
-
None
-
---
-
---
-
-
Moderate
-
None
Description of problem:
The "LiveMigratable" condition on the VMI is not populated during the scheduling phase of the virtual machine. If the VM is stuck in scheduling phase or if it takes more than a minute for the VM to move from Scheduling to Running, then the alert VMCannotBeEvicted is triggered since it get as an empty value for "LiveMigratable" condition.
VMI is in scheduling phase:
oc get vmi rhel9-silver-perch-55 NAME AGE PHASE IP NODENAME READY rhel9-silver-perch-55 21m Scheduling False
The virtual machine is added to "kubevirt_vmi_non_evictable":
$ oc exec virt-controller-6f88ff5dd9-t6chn -- curl -k https://127.0.0.1:8443/metrics 2>/dev/null |grep -i "rhel9-silver-perch-55"|egrep "kubevirt_vmi_info|evict" kubevirt_vmi_info{flavor="small",guest_os_kernel_release="",guest_os_machine="",guest_os_name="",guest_os_version_id="",instance_type="<none>",name="rhel9-silver-perch-55",namespace="nijin-cnv",node="",os="rhel9",phase="scheduling",preference="<none>",workload="server"} 1 kubevirt_vmi_non_evictable{name="rhel9-silver-perch-55",namespace="nijin-cnv",node=""} 1
And cluster is getting the following alert (notice the empty string in node name):
Eviction policy for VirtualMachine rhel9-silver-perch-55 in namespace Namespace NS nijin-cnv (on node ) is set to Live Migration but the VM is not migratable
Version-Release number of selected component (if applicable):
OpenShift Virtualization 4.17.4
How reproducible:
100%
Steps to Reproduce:
1. Create a VM and put an invalid node selector so that the VM will be stuck in scheduling phase. 2. Wait for a minute and look for the alert for VMCannotBeEvicted
Actual results:
VMCannotBeEvicted incorrectly triggered for scheduling VMs
Expected results:
Since the VM is not running yet and is not scheduled, OpenShift Virtualization shouldn't trigger the alert VMCannotBeEvicted
Additional info:
