Merge pull request #90046 from xenolinux/topology-spread-hcp

OSDOCS#12034: HCP Kubevirt topology spread constraint

Author: xenolinux

hosted_control_planes/hcp-manage/hcp-manage-virt.adoc

@@ -53,3 +53,10 @@ include::modules/hcp-virt-etcd-storage.adoc[leveloffset=+2]
 include::modules/hcp-virt-attach-nvidia-gpus.adoc[leveloffset=+1]
 
 include::modules/hcp-virt-attach-nvidia-gpus-np-api.adoc[leveloffset=+1]
+
+include::modules/hcp-topology-spread-constraint.adoc[leveloffset=+1]
+
+[role="_additional-resources"]
+.Additional resources
+
+* xref:../../nodes/scheduling/descheduler/nodes-descheduler-configuring.adoc#nodes-descheduler-installing_virt-enabling-descheduler-evictions[Installing the descheduler]

modules/hcp-topology-spread-constraint.adoc

@@ -0,0 +1,52 @@
+// Module included in the following assemblies:
+//
+// * hosted_control_planes/hcp-manage/hcp-manage-virt.adoc
+
+:_mod-docs-content-type: PROCEDURE
+[id="hcp-topology-spread-constraint_{context}"]
+= Spreading node pool VMs by using topologySpreadConstraint
+
+By default, KubeVirt virtual machines (VMs) created by a node pool are scheduled on any available nodes that have the capacity to run the VMs. By default, the `topologySpreadConstraint` constraint is set to schedule VMs on multiple nodes.
+
+In some scenarios, node pool VMs might run on the same node, which can cause availability issues. To avoid distribution of VMs on a single node, use the descheduler to continuously honour the `topologySpreadConstraint` constraint to spread VMs on multiple nodes.
+
+.Prerequisites
+
+* You installed the {descheduler-operator}. For more information, see "Installing the descheduler".
+
+.Procedure
+
+* Open the `KubeDescheduler` custom resource (CR) by entering the following command, and then modify the `KubeDescheduler` CR to use the `SoftTopologyAndDuplicates` profile so that you maintain the `topologySpreadConstraint` constraint settings.
++
+The `KubeDescheduler` CR named `cluster` runs in the `openshift-kube-descheduler-operator` namespace.
++
+[source,terminal]
+----
+$ oc edit kubedescheduler cluster -n openshift-kube-descheduler-operator
+----
++
+.Example `KubeDescheduler` configuration
+[source,yaml]
+----
+apiVersion: operator.openshift.io/v1
+kind: KubeDescheduler
+metadata:
+  name: cluster
+  namespace: openshift-kube-descheduler-operator
+spec:
+  mode: Automatic
+  managementState: Managed
+  deschedulingIntervalSeconds: 60 // <1>
+  profiles:
+  - SoftTopologyAndDuplicates // <2>
+  - EvictPodsWithPVC          // <3>
+  - EvictPodsWithLocalStorage // <4>
+  profileCustomizations:
+    devEnableEvictionsInBackground: true // <5>
+# ...
+----
+<1> Sets the number of seconds between the descheduler running cycles.
+<2> This profile evicts pods that follow the soft topology constraint: `whenUnsatisfiable: ScheduleAnyway`.
+<3> By default, the {descheduler-operator} prevents the pod eviction with persistent volume claims (PVCs). Use this profile to allow eviction of pods with PVCs.
+<4> By default, pods with local storage are not eligible for eviction. Use this profile to allow eviction of your VMs that use the local storage.
+<5> You must use this setting when performing a live migration so that the descheduler runs in the background during the migration process.