-
Task
-
Resolution: Unresolved
-
Minor
-
None
-
None
-
None
-
0.42
-
False
-
-
False
-
---
-
---
-
None
Description of problem:
In https://docs.openshift.com/container-platform/4.18/virt/nodes/virt-node-maintenance.html#eviction-strategies > Important section, it is mentioned as quoted below:
The default eviction strategy is LiveMigrate. A non-migratable VM with a LiveMigrate eviction strategy might prevent nodes from draining or block an infrastructure upgrade because the VM is not evicted from the node. This situation causes a migration to remain in a Pending or Scheduling state unless you shut down the VM manually.
But as I observed when a node with a running non-migratable VM with LiveMigrate evication strategy is drained, it is not actually causing migration to remain in a Pending or Scheduling state, rather it is not even creating a VirtualMachineInstanceMigration object at all. So IMO, we can remove the last sentence i.e.,"This situation causes a migration to remain in a Pending or Scheduling state unless you shut down the VM manually."
Version-Release number of selected component (if applicable):
brew.registry.redhat.io/rh-osbs/iib:888906
How reproducible:
consistently
Steps to Reproduce:
1. Create a non-migratable VM and start it. Can use attached manifest. 2. drain the worker node on which this VM is running
3. There is no VirtualMachineInstanceMigration object created at all, but in doc it is said the VirtualMachineInstanceMigration object will be created and will remain in Pending or Scheduling state, which seems not true.
Actual results:
Doc is incorrect
Expected results:
This sentence from doc needs to be removed: "This situation causes a migration to remain in a Pending or Scheduling state unless you shut down the VM manually."
Additional info:
This is verified on a 6 node s390x cluster in a PSI environment.
