CLOUDSTACK-9350: KVM-HA- Fix CheckOnHost for Local storage

- Also skip HA on VMs that are using local storage
alokhom · Apr 29, 2016 · 3b89cbe · 3b89cbe
1 parent e762e27
commit 3b89cbe
Show file tree

Hide file tree

Showing 7 changed files with 592 additions and 4 deletions.
diff --git a/plugins/hypervisors/kvm/src/com/cloud/ha/KVMInvestigator.java b/plugins/hypervisors/kvm/src/com/cloud/ha/KVMInvestigator.java
@@ -27,7 +27,11 @@
 import com.cloud.host.dao.HostDao;
 import com.cloud.hypervisor.Hypervisor;
 import com.cloud.resource.ResourceManager;
+import com.cloud.storage.Storage.StoragePoolType;
 import com.cloud.utils.component.AdapterBase;
+
+import org.apache.cloudstack.storage.datastore.db.PrimaryDataStoreDao;
+import org.apache.cloudstack.storage.datastore.db.StoragePoolVO;
 import org.apache.log4j.Logger;
 
 import javax.inject.Inject;
@@ -41,6 +45,8 @@ public class KVMInvestigator extends AdapterBase implements Investigator {
     AgentManager _agentMgr;
     @Inject
     ResourceManager _resourceMgr;
+    @Inject
+    PrimaryDataStoreDao _storagePoolDao;
 
     @Override
     public boolean isVmAlive(com.cloud.vm.VirtualMachine vm, Host host) throws UnknownVM {
@@ -60,6 +66,21 @@ public Status isAgentAlive(Host agent) {
         if (agent.getHypervisorType() != Hypervisor.HypervisorType.KVM && agent.getHypervisorType() != Hypervisor.HypervisorType.LXC) {
             return null;
         }
+
+        List<StoragePoolVO> clusterPools = _storagePoolDao.listPoolsByCluster(agent.getClusterId());
+        boolean hasNfs = false;
+        for (StoragePoolVO pool : clusterPools) {
+            if (pool.getPoolType() == StoragePoolType.NetworkFilesystem) {
+                hasNfs = true;
+                break;
+            }
+        }
+        if (!hasNfs) {
+            s_logger.warn(
+                    "Agent investigation was requested on host " + agent + ", but host does not support investigation because it has no NFS storage. Skipping investigation.");
+            return Status.Disconnected;
+        }
+
         Status hostStatus = null;
         Status neighbourStatus = null;
         CheckOnHostCommand cmd = new CheckOnHostCommand(agent);
@@ -78,7 +99,8 @@ public Status isAgentAlive(Host agent) {
 
         List<HostVO> neighbors = _resourceMgr.listHostsInClusterByStatus(agent.getClusterId(), Status.Up);
         for (HostVO neighbor : neighbors) {
-            if (neighbor.getId() == agent.getId() || (neighbor.getHypervisorType() != Hypervisor.HypervisorType.KVM && neighbor.getHypervisorType() != Hypervisor.HypervisorType.LXC)) {
+            if (neighbor.getId() == agent.getId()
+                    || (neighbor.getHypervisorType() != Hypervisor.HypervisorType.KVM && neighbor.getHypervisorType() != Hypervisor.HypervisorType.LXC)) {
                 continue;
             }
             s_logger.debug("Investigating host:" + agent.getId() + " via neighbouring host:" + neighbor.getId());

diff --git a/pom.xml b/pom.xml
@@ -921,6 +921,7 @@
               <exclude>**/.checkstyle</exclude>
               <exclude>scripts/installer/windows/acs_license.rtf</exclude>
               <exclude>**/*.md</exclude>
+              <exclude>test/integration/component/test_host_ha.sh</exclude>
             </excludes>
           </configuration>
         </plugin>

diff --git a/server/src/com/cloud/ha/HighAvailabilityManagerImpl.java b/server/src/com/cloud/ha/HighAvailabilityManagerImpl.java
@@ -62,6 +62,7 @@
 import com.cloud.hypervisor.Hypervisor.HypervisorType;
 import com.cloud.resource.ResourceManager;
 import com.cloud.server.ManagementServer;
+import com.cloud.service.ServiceOfferingVO;
 import com.cloud.service.dao.ServiceOfferingDao;
 import com.cloud.storage.StorageManager;
 import com.cloud.storage.dao.GuestOSCategoryDao;
@@ -264,6 +265,13 @@ public void scheduleRestartForVmsOnHost(final HostVO host, boolean investigate)
                 "Host [" + hostDesc + "] is down." + ((sb != null) ? sb.toString() : ""));
 
         for (VMInstanceVO vm : reorderedVMList) {
+            ServiceOfferingVO vmOffering = _serviceOfferingDao.findById(vm.getServiceOfferingId());
+            if (vmOffering.getUseLocalStorage()) {
+                if (s_logger.isDebugEnabled()){
+                    s_logger.debug("Skipping HA on vm " + vm + ", because it uses local storage. Its fate is tied to the host.");
+                }
+                continue;
+            }
             if (s_logger.isDebugEnabled()) {
                 s_logger.debug("Notifying HA Mgr of to restart vm " + vm.getId() + "-" + vm.getInstanceName());
             }

diff --git a/server/test/com/cloud/ha/HighAvailabilityManagerImplTest.java b/server/test/com/cloud/ha/HighAvailabilityManagerImplTest.java
@@ -59,6 +59,7 @@
 import com.cloud.hypervisor.Hypervisor.HypervisorType;
 import com.cloud.resource.ResourceManager;
 import com.cloud.server.ManagementServer;
+import com.cloud.service.ServiceOfferingVO;
 import com.cloud.service.dao.ServiceOfferingDao;
 import com.cloud.storage.StorageManager;
 import com.cloud.storage.dao.GuestOSCategoryDao;
@@ -195,6 +196,7 @@ public void scheduleRestartForVmsOnHostNonEmptyVMList() {
         Mockito.when(_dcDao.findById(Mockito.anyLong())).thenReturn(Mockito.mock(DataCenterVO.class));
         Mockito.when(_haDao.findPreviousHA(Mockito.anyLong())).thenReturn(Arrays.asList(Mockito.mock(HaWorkVO.class)));
         Mockito.when(_haDao.persist((HaWorkVO)Mockito.anyObject())).thenReturn(Mockito.mock(HaWorkVO.class));
+        Mockito.when(_serviceOfferingDao.findById(vm1.getServiceOfferingId())).thenReturn(Mockito.mock(ServiceOfferingVO.class));
 
         highAvailabilityManager.scheduleRestartForVmsOnHost(hostVO, true);
     }