From d5e5b48e906cc4a7e547a9c99e6117fa27d4cee1 Mon Sep 17 00:00:00 2001
From: Ranier Amorim <ranier.amorim@ratecnologia.inf.br>
Date: Mon, 4 May 2026 15:12:28 -0300
Subject: [PATCH] fix: skip VMs with host passthrough during DRS plan
 generation

When iterating over VMs in getBestMigration, the call to
listHostsForMigrationOfVM throws InvalidParameterValueException
for VMs using a vGPU passthrough profile. This exception was not
caught inside the loop, causing it to propagate up and abort the
entire DRS plan generation for the cluster, leaving all other
eligible VMs without a migration plan.

Fix wraps the listHostsForMigrationOfVM call in a try-catch block.
When the exception is caught, the VM is skipped with a debug log
message and the loop continues evaluating the remaining VMs.

Also adds unit test testGetBestMigrationSkipsPassthroughVm to verify
that a VM throwing InvalidParameterValueException is skipped while
other eligible VMs are still considered for migration.

Fixes: #13098
---
 .../cluster/ClusterDrsServiceImpl.java        | 12 ++++--
 .../cluster/ClusterDrsServiceImplTest.java    | 43 +++++++++++++++++++
 2 files changed, 52 insertions(+), 3 deletions(-)

diff --git a/server/src/main/java/org/apache/cloudstack/cluster/ClusterDrsServiceImpl.java b/server/src/main/java/org/apache/cloudstack/cluster/ClusterDrsServiceImpl.java
index a662d47d4541..dcde1e219754 100644
--- a/server/src/main/java/org/apache/cloudstack/cluster/ClusterDrsServiceImpl.java
+++ b/server/src/main/java/org/apache/cloudstack/cluster/ClusterDrsServiceImpl.java
@@ -449,9 +449,15 @@ Pair<VirtualMachine, Host> getBestMigration(Cluster cluster, ClusterDrsAlgorithm
             ) {
                 continue;
             }
-            Ternary<Pair<List<? extends Host>, Integer>, List<? extends Host>, Map<Host, Boolean>> hostsForMigrationOfVM = managementServer
-                    .listHostsForMigrationOfVM(
-                            vm, 0L, 500L, null, vmList);
+            Ternary<Pair<List<? extends Host>, Integer>, List<? extends Host>, Map<Host, Boolean>> hostsForMigrationOfVM;
+            try {
+                hostsForMigrationOfVM = managementServer
+                        .listHostsForMigrationOfVM(
+                                vm, 0L, 500L, null, vmList);
+            } catch (InvalidParameterValueException e) {
+                logger.debug("Skipping VM {} for DRS, unsupported operation: {}", vm, e.getMessage());
+                continue;
+            }
             List<? extends Host> compatibleDestinationHosts = hostsForMigrationOfVM.first().first();
             List<? extends Host> suitableDestinationHosts = hostsForMigrationOfVM.second();
 
diff --git a/server/src/test/java/org/apache/cloudstack/cluster/ClusterDrsServiceImplTest.java b/server/src/test/java/org/apache/cloudstack/cluster/ClusterDrsServiceImplTest.java
index 81aac9e4b54d..381fb08b08e1 100644
--- a/server/src/test/java/org/apache/cloudstack/cluster/ClusterDrsServiceImplTest.java
+++ b/server/src/test/java/org/apache/cloudstack/cluster/ClusterDrsServiceImplTest.java
@@ -404,6 +404,49 @@ public void testGetBestMigration() throws ConfigurationException {
         assertEquals(vm1, bestMigration.first());
     }
 
+    @Test
+    public void testGetBestMigrationSkipsPassthroughVm() throws ConfigurationException {
+        ClusterVO cluster = Mockito.mock(ClusterVO.class);
+        Mockito.when(cluster.getId()).thenReturn(1L);
+
+        HostVO destHost = Mockito.mock(HostVO.class);
+        Mockito.when(destHost.getClusterId()).thenReturn(1L);
+
+        VMInstanceVO vmPassthrough = Mockito.mock(VMInstanceVO.class);
+        Mockito.when(vmPassthrough.getId()).thenReturn(1L);
+        Mockito.when(vmPassthrough.getType()).thenReturn(VirtualMachine.Type.User);
+        Mockito.when(vmPassthrough.getState()).thenReturn(VirtualMachine.State.Running);
+        Mockito.when(vmPassthrough.getDetails()).thenReturn(Collections.emptyMap());
+
+        VMInstanceVO vmNormal = Mockito.mock(VMInstanceVO.class);
+        Mockito.when(vmNormal.getId()).thenReturn(2L);
+        Mockito.when(vmNormal.getType()).thenReturn(VirtualMachine.Type.User);
+        Mockito.when(vmNormal.getState()).thenReturn(VirtualMachine.State.Running);
+        Mockito.when(vmNormal.getDetails()).thenReturn(Collections.emptyMap());
+
+        List<VirtualMachine> vmList = new ArrayList<>();
+        vmList.add(vmPassthrough);
+        vmList.add(vmNormal);
+
+        ServiceOffering serviceOffering = Mockito.mock(ServiceOffering.class);
+        Map<Long, ServiceOffering> vmIdServiceOfferingMap = new HashMap<>();
+        vmIdServiceOfferingMap.put(vmPassthrough.getId(), serviceOffering);
+        vmIdServiceOfferingMap.put(vmNormal.getId(), serviceOffering);
+
+        Mockito.when(managementServer.listHostsForMigrationOfVM(vmPassthrough, 0L, 500L, null, vmList))
+                .thenThrow(new InvalidParameterValueException("Unsupported operation, VM uses host passthrough, cannot migrate"));
+        Mockito.when(managementServer.listHostsForMigrationOfVM(vmNormal, 0L, 500L, null, vmList)).thenReturn(
+                new Ternary<>(new Pair<>(List.of(destHost), 1), List.of(destHost), Map.of(destHost, false)));
+        Mockito.when(balancedAlgorithm.getMetrics(cluster, vmNormal, serviceOffering, destHost, new HashMap<>(),
+                new HashMap<>(), false)).thenReturn(new Ternary<>(1.0, 0.5, 1.5));
+
+        Pair<VirtualMachine, Host> bestMigration = clusterDrsService.getBestMigration(cluster, balancedAlgorithm,
+                vmList, vmIdServiceOfferingMap, new HashMap<>(), new HashMap<>());
+
+        assertEquals(vmNormal, bestMigration.first());
+        assertEquals(destHost, bestMigration.second());
+    }
+
     @Test
     public void testGetBestMigrationDifferentCluster() throws ConfigurationException {
         ClusterVO cluster = Mockito.mock(ClusterVO.class);