core: validate amount of sockets equal to that of numa nodes

When running a VM with 'Resize and Pin' CPU pinning policy, the algorithm expects the host to have the same number of sockets and NUMA nodes. That way the algorithm fits the topology and leaves the first core on each socket free to the host. This patch filters out hosts that don't have an equal number of sockets and NUMA nodes or the host doesn't have NUMA support in the NUMA filter policy when trying to run a VM with CPU pinning policy `Resize and Pin`. Change-Id: Icdc1ac910eb733f6edd861d0505fb90a581224b5 Bug-Url: https://bugzilla.redhat.com/1955388 Bug-Url: https://bugzilla.redhat.com/2056950 Signed-off-by: Liran Rotenberg <lrotenbe@redhat.com>
oVirt · Jun 22, 2022 · 437a831 · 437a831
1 parent 513d016
commit 437a831
Show file tree

Hide file tree

Showing 4 changed files with 16 additions and 3 deletions.
diff --git a/...es/bll/src/main/java/org/ovirt/engine/core/bll/scheduling/policyunits/NumaPolicyUnit.java b/...es/bll/src/main/java/org/ovirt/engine/core/bll/scheduling/policyunits/NumaPolicyUnit.java
@@ -17,6 +17,7 @@
 import org.ovirt.engine.core.common.scheduling.PolicyUnit;
 import org.ovirt.engine.core.common.scheduling.PolicyUnitType;
 import org.ovirt.engine.core.common.utils.NumaPinningHelper;
+import org.ovirt.engine.core.common.utils.VmCpuCountHelper;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
@@ -37,8 +38,8 @@ public NumaPolicyUnit(PolicyUnit policyUnit,
     @Override
     public List<VDS> filter(SchedulingContext context, List<VDS> hosts, List<VM> vmGroup, PerHostMessages messages) {
         boolean vmNumaPinned = vmGroup.stream()
-                .flatMap(vm -> vm.getvNumaNodeList().stream())
-                .anyMatch(node -> !node.getVdsNumaNodeList().isEmpty());
+                .anyMatch(vm -> vm.getvNumaNodeList().stream().anyMatch(node -> !node.getVdsNumaNodeList().isEmpty())
+                        || VmCpuCountHelper.isResizeAndPinPolicy(vm));
 
         // If no VM numa node is pinned, all hosts are accepted.
         //
@@ -48,7 +49,7 @@ public List<VDS> filter(SchedulingContext context, List<VDS> hosts, List<VM> vmG
         }
 
         List<VM> vmsToCheck = vmGroup.stream()
-                .filter(vm -> !vm.getvNumaNodeList().isEmpty())
+                .filter(vm -> !vm.getvNumaNodeList().isEmpty() || VmCpuCountHelper.isResizeAndPinPolicy(vm))
                 .collect(Collectors.toList());
 
         if (vmsToCheck.isEmpty()) {
@@ -57,6 +58,14 @@ public List<VDS> filter(SchedulingContext context, List<VDS> hosts, List<VM> vmG
 
         List<VDS> result = new ArrayList<>();
         for (VDS host: hosts) {
+            if (vmsToCheck.stream().anyMatch(VmCpuCountHelper::isResizeAndPinPolicy) &&
+                    host.getCpuSockets() != host.getNumaNodeList().size()) {
+                log.debug("Host '{}' is not qualified for the 'Resize and Pin' CPU pinning policy," +
+                        " number of sockets is not equal to number of NUMA nodes", host.getName());
+                messages.addMessage(host.getId(), EngineMessage.VAR__DETAIL__SOCKETS_UNEQUAL_NUMA.toString());
+                continue;
+            }
+
             List<VM> vmsToCheckOnHost = new ArrayList<>(vmsToCheck.size());
             boolean skipHostCheck = true;
             for (VM vm : vmsToCheck) {

diff --git a/...nager/modules/common/src/main/java/org/ovirt/engine/core/common/errors/EngineMessage.java b/...nager/modules/common/src/main/java/org/ovirt/engine/core/common/errors/EngineMessage.java
@@ -1554,6 +1554,7 @@ public enum EngineMessage {
     VAR__DETAIL__NOT_ENOUGH_HUGE_PAGES,
     VAR__DETAIL__NUMA_NOT_SUPPORTED,
     VAR__DETAIL__NOT_MEMORY_PINNED_NUMA,
+    VAR__DETAIL__SOCKETS_UNEQUAL_NUMA,
     VAR__DETAIL__NOT_ENOUGH_CORES,
     VAR__DETAIL__NOT_ENOUGH_CORES_PER_SOCKET_FOR_RESIZE_AND_PIN,
     VAR__DETAIL__NOT_ENOUGH_THREADS_PER_CORE_FOR_DEDICATED,

diff --git a/frontend/webadmin/modules/frontend/src/main/java/org/ovirt/engine/ui/frontend/AppErrors.java b/frontend/webadmin/modules/frontend/src/main/java/org/ovirt/engine/ui/frontend/AppErrors.java
@@ -2705,6 +2705,8 @@ public interface AppErrors extends ConstantsWithLookup {
 
     String VAR__DETAIL__NOT_MEMORY_PINNED_NUMA();
 
+    String VAR__DETAIL__SOCKETS_UNEQUAL_NUMA();
+
     String VAR__DETAIL__NOT_ENOUGH_CORES();
 
     String VAR__DETAIL__NOT_ENOUGH_CORES_PER_SOCKET_FOR_RESIZE_AND_PIN();

diff --git a/...min/modules/frontend/src/main/resources/org/ovirt/engine/ui/frontend/AppErrors.properties b/...min/modules/frontend/src/main/resources/org/ovirt/engine/ui/frontend/AppErrors.properties
@@ -1294,6 +1294,7 @@ VAR__DETAIL__HE_HOST_NOT_POSITIVE_SCORE=$detailMessage it doesn't have a positiv
 VAR__DETAIL__HE_HOST_LOW_SCORE=$detailMessage it has much lower Hosted Engine score than the current host.
 VAR__DETAIL__NUMA_NOT_SUPPORTED=$detailMessage does not support NUMA
 VAR__DETAIL__NOT_MEMORY_PINNED_NUMA=$detailMessage cannot accommodate memory of VM's pinned virtual NUMA nodes within host's physical NUMA nodes.
+VAR__DETAIL__SOCKETS_UNEQUAL_NUMA=$detailMessage cannot apply the CPU pinning policy 'Resize and Pin' on a host when the number of CPU sockets is not equal to the number of NUMA nodes
 VAR__DETAIL__NO_SUITABLE_VF=$detailMessage there are no free virtual functions which are suitable for virtual nic(s) ${vnicNames}. A virtual function is considered as suitable if the VF's configuration of its physical function contains the virtual nic's network/network label
 VAR__DETAIL__NOT_PINNED_TO_HOST=$detailMessage the host is not present in the pinning list.
 VAR__DETAIL__MISSING_VDS_HOOK=$detailMessage the host does not contain required VDS hook ${vdsHook}