Merge pull request #10 from shapeblue/kvm-ha

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

Author: John Burwell

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.ejb.Local;
@@ -43,6 +47,8 @@ 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) {
@@ -58,6 +64,20 @@ 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);

plugins/hypervisors/kvm/src/com/cloud/hypervisor/kvm/resource/LibvirtComputingResource.java

@@ -1744,6 +1744,7 @@ protected Answer execute(CheckOnHostCommand cmd) {
             if (result) {
                 return new Answer(cmd, false, "Heart is still beating...");
             } else {
+                s_logger.warn("Heartbeat failed for : " + cmd.getHost().getPrivateNetwork().getIp().toString());
                 return new Answer(cmd);
             }
         } catch (InterruptedException e) {

pom.xml

@@ -828,6 +828,7 @@
               <exclude>tools/ngui/static/js/lib/*</exclude>
               <exclude>**/.checkstyle</exclude>
               <exclude>scripts/installer/windows/acs_license.rtf</exclude>
+              <exclude>test/integration/component/test_host_ha.sh</exclude>
             </excludes>
           </configuration>
         </plugin>

server/src/com/cloud/ha/HighAvailabilityManagerImpl.java

@@ -64,6 +64,7 @@
 import com.cloud.resource.ResourceManager;
 import com.cloud.server.ManagementServer;
 import com.cloud.service.dao.ServiceOfferingDao;
+import com.cloud.service.ServiceOfferingVO;
 import com.cloud.storage.StorageManager;
 import com.cloud.storage.dao.GuestOSCategoryDao;
 import com.cloud.storage.dao.GuestOSDao;
@@ -267,6 +268,13 @@ public void scheduleRestartForVmsOnHost(final HostVO host, boolean investigate)
 
         if (vms != null) {
             for (VMInstanceVO vm : vms) {
+                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());
                 }