Live scaling for VMs with fixed service offerings on KVM

Author: bernardodemarco

api/src/main/java/com/cloud/agent/api/to/VirtualMachineTO.java

@@ -51,6 +51,7 @@ public class VirtualMachineTO {
 
     private long minRam;
     private long maxRam;
+    private long requestedRam;
     private String hostName;
     private String arch;
     private String os;
@@ -207,15 +208,20 @@ public long getMinRam() {
         return minRam;
     }
 
-    public void setRam(long minRam, long maxRam) {
+    public void setRam(long minRam, long maxRam, long requestedRam) {
         this.minRam = minRam;
         this.maxRam = maxRam;
+        this.requestedRam = requestedRam;
     }
 
     public long getMaxRam() {
         return maxRam;
     }
 
+    public long getRequestedRam() {
+        return requestedRam;
+    }
+
     public String getHostName() {
         return hostName;
     }

core/src/main/java/com/cloud/agent/api/ScaleVmCommand.java

@@ -30,6 +30,7 @@ public class ScaleVmCommand extends Command {
     Integer maxSpeed;
     long minRam;
     long maxRam;
+    private boolean limitCpuUseChange;
 
     public VirtualMachineTO getVm() {
         return vm;
@@ -43,7 +44,7 @@ public int getCpus() {
         return cpus;
     }
 
-    public ScaleVmCommand(String vmName, int cpus, Integer minSpeed, Integer maxSpeed, long minRam, long maxRam, boolean limitCpuUse) {
+    public ScaleVmCommand(String vmName, int cpus, Integer minSpeed, Integer maxSpeed, long minRam, long maxRam, boolean limitCpuUse, Double cpuQuotaPercentage, boolean limitCpuUseChange) {
         super();
         this.vmName = vmName;
         this.cpus = cpus;
@@ -52,6 +53,8 @@ public ScaleVmCommand(String vmName, int cpus, Integer minSpeed, Integer maxSpee
         this.minRam = minRam;
         this.maxRam = maxRam;
         this.vm = new VirtualMachineTO(1L, vmName, null, cpus, minSpeed, maxSpeed, minRam, maxRam, null, null, false, limitCpuUse, null);
+        this.vm.setCpuQuotaPercentage(cpuQuotaPercentage);
+        this.limitCpuUseChange = limitCpuUseChange;
     }
 
     public void setCpus(int cpus) {
@@ -102,6 +105,10 @@ public VirtualMachineTO getVirtualMachine() {
         return vm;
     }
 
+    public boolean getLimitCpuUseChange() {
+        return limitCpuUseChange;
+    }
+
     @Override
     public boolean executeInSequence() {
         return true;

engine/components-api/src/main/java/com/cloud/capacity/CapacityManager.java

@@ -133,6 +133,20 @@ public interface CapacityManager {
             "capacity.calculate.workers", "1",
             "Number of worker threads to be used for capacities calculation", true);
 
+    ConfigKey<Integer> KvmMemoryDynamicScalingCapacity = new ConfigKey<>(ConfigKey.CATEGORY_ADVANCED,
+            Integer.class, "kvm.memory.dynamic.scaling.capacity", "0",
+            "Defines the maximum memory capacity in MiB for which VMs can be dynamically scaled to with KVM. " +
+                    "The 'kvm.memory.dynamic.scaling.capacity' setting's value will be used to define the value of the " +
+                    "'<maxMemory />' element of domain XMLs. If it is set to a value less than or equal to '0', then the host's memory capacity will be considered.",
+            true, ConfigKey.Scope.Cluster);
+
+    ConfigKey<Integer> KvmCpuDynamicScalingCapacity = new ConfigKey<>(ConfigKey.CATEGORY_ADVANCED,
+            Integer.class, "kvm.cpu.dynamic.scaling.capacity", "0",
+            "Defines the maximum vCPU capacity for which VMs can be dynamically scaled to with KVM. " +
+                    "The 'kvm.cpu.dynamic.scaling.capacity' setting's value will be used to define the value of the " +
+                    "'<vcpu />' element of domain XMLs. If it is set to a value less than or equal to '0', then the host's CPU cores capacity will be considered.",
+            true, ConfigKey.Scope.Cluster);
+
     public boolean releaseVmCapacity(VirtualMachine vm, boolean moveFromReserved, boolean moveToReservered, Long hostId);
 
     void allocateVmCapacity(VirtualMachine vm, boolean fromLastHost);

engine/orchestration/src/main/java/com/cloud/vm/VirtualMachineManagerImpl.java

@@ -50,6 +50,7 @@
 import javax.naming.ConfigurationException;
 import javax.persistence.EntityExistsException;
 
+import com.cloud.hypervisor.KVMGuru;
 import org.apache.cloudstack.affinity.dao.AffinityGroupVMMapDao;
 import org.apache.cloudstack.annotation.AnnotationService;
 import org.apache.cloudstack.annotation.dao.AnnotationDao;
@@ -5180,7 +5181,7 @@ public VMInstanceVO reConfigureVm(final String vmUuid, final ServiceOffering old
             try {
                 result = retrieveResultFromJobOutcomeAndThrowExceptionIfNeeded(outcome);
             } catch (Exception ex) {
-                throw new RuntimeException("Unhandled exception", ex);
+                throw new RuntimeException("Unable to reconfigure VM.", ex);
             }
 
             if (result != null) {
@@ -5193,22 +5194,29 @@ public VMInstanceVO reConfigureVm(final String vmUuid, final ServiceOffering old
 
     private VMInstanceVO orchestrateReConfigureVm(String vmUuid, ServiceOffering oldServiceOffering, ServiceOffering newServiceOffering,
                                                   boolean reconfiguringOnExistingHost) throws ResourceUnavailableException, ConcurrentOperationException {
-        final VMInstanceVO vm = _vmDao.findByUuid(vmUuid);
+        VMInstanceVO vm = _vmDao.findByUuid(vmUuid);
 
         HostVO hostVo = _hostDao.findById(vm.getHostId());
 
-        Long clustedId = hostVo.getClusterId();
-        Float memoryOvercommitRatio = CapacityManager.MemOverprovisioningFactor.valueIn(clustedId);
-        Float cpuOvercommitRatio = CapacityManager.CpuOverprovisioningFactor.valueIn(clustedId);
-        boolean divideMemoryByOverprovisioning = HypervisorGuruBase.VmMinMemoryEqualsMemoryDividedByMemOverprovisioningFactor.valueIn(clustedId);
-        boolean divideCpuByOverprovisioning = HypervisorGuruBase.VmMinCpuSpeedEqualsCpuSpeedDividedByCpuOverprovisioningFactor.valueIn(clustedId);
+        Long clusterId = hostVo.getClusterId();
+        Float memoryOvercommitRatio = CapacityManager.MemOverprovisioningFactor.valueIn(clusterId);
+        Float cpuOvercommitRatio = CapacityManager.CpuOverprovisioningFactor.valueIn(clusterId);
+        boolean divideMemoryByOverprovisioning = HypervisorGuruBase.VmMinMemoryEqualsMemoryDividedByMemOverprovisioningFactor.valueIn(clusterId);
+        boolean divideCpuByOverprovisioning = HypervisorGuruBase.VmMinCpuSpeedEqualsCpuSpeedDividedByCpuOverprovisioningFactor.valueIn(clusterId);
 
         int minMemory = (int)(newServiceOffering.getRamSize() / (divideMemoryByOverprovisioning ? memoryOvercommitRatio : 1));
         int minSpeed = (int)(newServiceOffering.getSpeed() / (divideCpuByOverprovisioning ? cpuOvercommitRatio : 1));
 
-        ScaleVmCommand scaleVmCommand =
-                new ScaleVmCommand(vm.getInstanceName(), newServiceOffering.getCpu(), minSpeed,
-                        newServiceOffering.getSpeed(), minMemory * 1024L * 1024L, newServiceOffering.getRamSize() * 1024L * 1024L, newServiceOffering.getLimitCpuUse());
+        Double cpuQuotaPercentage = null;
+        if (newServiceOffering.getLimitCpuUse() && vm.getHypervisorType().equals(HypervisorType.KVM)) {
+            KVMGuru kvmGuru = (KVMGuru) _hvGuruMgr.getGuru(vm.getHypervisorType());
+            cpuQuotaPercentage = kvmGuru.getCpuQuotaPercentage(minSpeed, hostVo.getSpeed());
+        }
+
+        boolean limitCpuUseChange = oldServiceOffering.getLimitCpuUse() != newServiceOffering.getLimitCpuUse();
+        ScaleVmCommand scaleVmCommand = new ScaleVmCommand(vm.getInstanceName(), newServiceOffering.getCpu(), minSpeed, newServiceOffering.getSpeed(),
+                minMemory * 1024L * 1024L, newServiceOffering.getRamSize() * 1024L * 1024L,
+                newServiceOffering.getLimitCpuUse(), cpuQuotaPercentage, limitCpuUseChange);
 
         scaleVmCommand.getVirtualMachine().setId(vm.getId());
         scaleVmCommand.getVirtualMachine().setUuid(vm.getUuid());
@@ -5237,16 +5245,20 @@ private VMInstanceVO orchestrateReConfigureVm(String vmUuid, ServiceOffering old
                 throw new CloudRuntimeException("Unable to scale vm due to " + (reconfigureAnswer == null ? "" : reconfigureAnswer.getDetails()));
             }
 
-            upgradeVmDb(vm.getId(), newServiceOffering, oldServiceOffering);
+            if (reconfiguringOnExistingHost) {
+                _capacityMgr.releaseVmCapacity(vm, false, false, vm.getHostId());
+            }
+
+            boolean vmUpgraded = upgradeVmDb(vm.getId(), newServiceOffering, oldServiceOffering);
+            if (vmUpgraded) {
+                vm = _vmDao.findById(vm.getId());
+            }
 
             if (vm.getType().equals(VirtualMachine.Type.User)) {
                 _userVmMgr.generateUsageEvent(vm, vm.isDisplayVm(), EventTypes.EVENT_VM_DYNAMIC_SCALE);
             }
 
             if (reconfiguringOnExistingHost) {
-                vm.setServiceOfferingId(oldServiceOffering.getId());
-                _capacityMgr.releaseVmCapacity(vm, false, false, vm.getHostId());
-                vm.setServiceOfferingId(newServiceOffering.getId());
                 _capacityMgr.allocateVmCapacity(vm, false);
             }
 

engine/schema/src/main/resources/META-INF/db/schema-42210to42300.sql

@@ -117,3 +117,21 @@ CALL `cloud`.`IDEMPOTENT_ADD_COLUMN`('cloud.vpc_offerings','conserve_mode', 'tin
 
 --- Disable/enable NICs
 CALL `cloud`.`IDEMPOTENT_ADD_COLUMN`('cloud.nics','enabled', 'TINYINT(1) NOT NULL DEFAULT 1 COMMENT ''Indicates whether the NIC is enabled or not'' ');
+
+-- Creates the 'kvm.memory.dynamic.scaling.capacity' and, for already active ACS environments,
+-- initializes it with the value of the setting 'vm.serviceoffering.ram.size.max'
+INSERT INTO `cloud`.`configuration` (`category`, `instance`, `component`, `name`, `value`, `default_value`, `updated`, `scope`, `is_dynamic`, `group_id`, `subgroup_id`, `display_text`, `description`)
+SELECT 'Advanced', 'DEFAULT', 'CapacityManager', 'kvm.memory.dynamic.scaling.capacity', `cfg`.`value`, 0, NULL, 4, 1, 6, 27,
+       'KVM memory dynamic scaling capacity', 'Defines the maximum memory capacity in MiB for which VMs can be dynamically scaled to with KVM. The ''kvm.memory.dynamic.scaling.capacity'' setting''s value will be used to define the value of the ''<maxMemory />'' element of domain XMLs. If it is set to a value less than or equal to ''0'', then the host''s memory capacity will be considered.'
+FROM `cloud`.`configuration` `cfg`
+WHERE NOT EXISTS (SELECT 1 FROM `cloud`.`configuration` WHERE `name` = 'kvm.memory.dynamic.scaling.capacity')
+  AND `cfg`.`name` = 'vm.serviceoffering.ram.size.max';
+
+-- Creates the 'kvm.cpu.dynamic.scaling.capacity' and, for already active ACS environments,
+-- initializes it with the value of the setting 'vm.serviceoffering.cpu.cores.max'
+INSERT INTO `cloud`.`configuration` (`category`, `instance`, `component`, `name`, `value`, `default_value`, `updated`, `scope`, `is_dynamic`, `group_id`, `subgroup_id`, `display_text`, `description`)
+SELECT 'Advanced', 'DEFAULT', 'CapacityManager', 'kvm.cpu.dynamic.scaling.capacity', `cfg`.`value`, 0, NULL, 4, 1, 6, 27,
+       'KVM CPU dynamic scaling capacity', 'Defines the maximum vCPU capacity for which VMs can be dynamically scaled to with KVM. The ''kvm.cpu.dynamic.scaling.capacity'' setting''s value will be used to define the value of the ''<vcpu />'' element of domain XMLs. If it is set to a value less than or equal to ''0'', then the host''s CPU cores capacity will be considered.'
+FROM `cloud`.`configuration` `cfg`
+WHERE NOT EXISTS (SELECT 1 FROM `cloud`.`configuration` WHERE `name` = 'kvm.cpu.dynamic.scaling.capacity')
+  AND `cfg`.`name` = 'vm.serviceoffering.cpu.cores.max';