server: investigate pending HA work when executing in new MS session (#10167)

For HA work items that are created for host state change, checks must be
done when execution is called in a new management server session.

A new column, reason, has been added in cloud.op_ha_work table to track
the reason for HA work.

When HighAvailabilityManager starts it finds and puts all pending HA
work items in Investigating state. During execution of the HA work if it
is found in investigating state, checks are done to verify if the work
is still valid. If the jobs is found to be invalid it is cancelled.

Signed-off-by: Abhishek Kumar <[email protected]>

Author: shwstppr

engine/components-api/src/main/java/com/cloud/ha/HighAvailabilityManager.java

@@ -84,6 +84,13 @@ public enum WorkType {
         HA;         // Restart a VM.
     }
 
+    enum ReasonType {
+        Unknown,
+        HostMaintenance,
+        HostDown,
+        HostDegraded;
+    }
+
     enum Step {
         Scheduled, Investigating, Fencing, Stopping, Restarting, Migrating, Cancelled, Done, Error,
     }
@@ -92,7 +99,7 @@ enum Step {
      * Investigate why a host has disconnected and migrate the VMs on it
      * if necessary.
      *
-     * @param host - the host that has disconnected.
+     * @param hostId - the id of the host that has disconnected.
      */
     Status investigate(long hostId);
 
@@ -109,17 +116,19 @@ enum Step {
      * @param investigate must be investigated before we do anything with this vm.
      */
     void scheduleRestart(VMInstanceVO vm, boolean investigate);
+    void scheduleRestart(VMInstanceVO vm, boolean investigate, ReasonType reasonType);
 
     void cancelDestroy(VMInstanceVO vm, Long hostId);
 
-    boolean scheduleDestroy(VMInstanceVO vm, long hostId);
+    boolean scheduleDestroy(VMInstanceVO vm, long hostId, ReasonType reasonType);
 
     /**
      * Schedule restarts for all vms running on the host.
      * @param host host.
-     * @param investigate TODO
+     * @param investigate whether to investigate
+     * @param reasonType reason for HA work
      */
-    void scheduleRestartForVmsOnHost(HostVO host, boolean investigate);
+    void scheduleRestartForVmsOnHost(HostVO host, boolean investigate, ReasonType reasonType);
 
     /**
      * Schedule the vm for migration.
@@ -128,6 +137,7 @@ enum Step {
      * @return true if schedule worked.
      */
     boolean scheduleMigration(VMInstanceVO vm);
+    boolean scheduleMigration(VMInstanceVO vm, ReasonType reasonType);
 
     List<VMInstanceVO> findTakenMigrationWork();
 
@@ -140,10 +150,11 @@ enum Step {
      * 3. Check if a VM has been stopped: WorkType.CheckStop
      *
      * @param vm virtual machine to stop.
-     * @param host host the virtual machine is on.
+     * @param hostId the id of the host the virtual machine is on.
      * @param type which type of stop is requested.
      */
     boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type);
+    boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type, ReasonType reasonType);
 
     void cancelScheduledMigrations(HostVO host);
 

engine/orchestration/src/main/java/com/cloud/agent/manager/AgentManagerImpl.java

@@ -989,7 +989,7 @@ protected boolean handleDisconnectWithInvestigation(final AgentAttache attache,
         handleDisconnectWithoutInvestigation(attache, event, true, true);
         host = _hostDao.findById(hostId); // Maybe the host magically reappeared?
         if (host != null && host.getStatus() == Status.Down) {
-            _haMgr.scheduleRestartForVmsOnHost(host, true);
+            _haMgr.scheduleRestartForVmsOnHost(host, true, HighAvailabilityManager.ReasonType.HostDown);
         }
         return true;
     }

engine/schema/src/main/resources/META-INF/db/schema-42000to42010.sql

@@ -35,3 +35,6 @@ CALL `cloud`.`IDEMPOTENT_ADD_COLUMN`('cloud.volumes', 'last_id', 'bigint(20) uns
 
 -- Add used_iops column to support IOPS data in storage stats
 CALL `cloud`.`IDEMPOTENT_ADD_COLUMN`('cloud.storage_pool', 'used_iops', 'bigint unsigned DEFAULT NULL COMMENT "IOPS currently in use for this storage pool" ');
+
+-- Add reason column for op_ha_work
+CALL `cloud`.`IDEMPOTENT_ADD_COLUMN`('cloud.op_ha_work', 'reason', 'varchar(32) DEFAULT NULL COMMENT "Reason for the HA work"');

server/src/main/java/com/cloud/ha/HaWorkVO.java

@@ -86,6 +86,10 @@ public class HaWorkVO implements InternalIdentity {
     @Column(name = "tried")
     int timesTried;
 
+    @Column(name = "reason")
+    @Enumerated(value = EnumType.STRING)
+    private HighAvailabilityManager.ReasonType reasonType;
+
     protected HaWorkVO() {
     }
 
@@ -179,7 +183,7 @@ public void setPreviousState(State state) {
     }
 
     public HaWorkVO(final long instanceId, final VirtualMachine.Type type, final WorkType workType, final Step step, final long hostId, final State previousState,
-            final int timesTried, final long updated) {
+            final int timesTried, final long updated, HighAvailabilityManager.ReasonType reasonType) {
         this.workType = workType;
         this.type = type;
         this.instanceId = instanceId;
@@ -191,6 +195,7 @@ public HaWorkVO(final long instanceId, final VirtualMachine.Type type, final Wor
         this.step = step;
         this.timeToTry = System.currentTimeMillis() >> 10;
         this.updateTime = updated;
+        this.reasonType = reasonType;
     }
 
     @Override
@@ -207,4 +212,12 @@ public String toString() {
             .append("]")
             .toString();
     }
+
+    public HighAvailabilityManager.ReasonType getReasonType() {
+        return reasonType;
+    }
+
+    public void setReasonType(HighAvailabilityManager.ReasonType reasonType) {
+        this.reasonType = reasonType;
+    }
 }

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

@@ -19,6 +19,7 @@
 import static org.apache.cloudstack.framework.config.ConfigKey.Scope.Zone;
 
 import java.util.ArrayList;
+import java.util.Arrays;
 import java.util.Date;
 import java.util.HashMap;
 import java.util.List;
@@ -43,6 +44,7 @@
 import org.apache.cloudstack.managed.context.ManagedContext;
 import org.apache.cloudstack.managed.context.ManagedContextRunnable;
 import org.apache.cloudstack.management.ManagementServerHost;
+import org.apache.logging.log4j.ThreadContext;
 
 import com.cloud.agent.AgentManager;
 import com.cloud.alert.AlertManager;
@@ -90,7 +92,6 @@
 import com.cloud.vm.VirtualMachineManager;
 import com.cloud.vm.VirtualMachineProfile;
 import com.cloud.vm.dao.VMInstanceDao;
-import org.apache.logging.log4j.ThreadContext;
 
 /**
  * HighAvailabilityManagerImpl coordinates the HA process. VMs are registered with the HA Manager for HA. The request is stored
@@ -133,6 +134,9 @@ public class HighAvailabilityManagerImpl extends ManagerBase implements Configur
     protected static ConfigKey<Boolean> VmHaAlertsEnabled = new ConfigKey<>("Advanced", Boolean.class, "vm.ha.alerts.enabled", "true",
             "Enable/Disable alerts for the VM HA operations, it is enabled by default.", true, Zone);
 
+    protected static final List<ReasonType> CancellableWorkReasonTypes =
+            Arrays.asList(ReasonType.HostMaintenance, ReasonType.HostDown, ReasonType.HostDegraded);
+
     WorkerThread[] _workers;
     boolean _stopped;
     long _timeToSleep;
@@ -269,8 +273,7 @@ public Status investigate(final long hostId) {
     }
 
     @Override
-    public void scheduleRestartForVmsOnHost(final HostVO host, boolean investigate) {
-
+    public void scheduleRestartForVmsOnHost(final HostVO host, boolean investigate, ReasonType reasonType) {
         if (host.getType() != Host.Type.Routing) {
             return;
         }
@@ -337,12 +340,12 @@ public void scheduleRestartForVmsOnHost(final HostVO host, boolean investigate)
                 logger.debug("VM {} is not on down host {} it is on other host {} VM HA is done", vm, host, hostId);
                 continue;
             }
-            scheduleRestart(vm, investigate);
+            scheduleRestart(vm, investigate, reasonType);
         }
     }
 
     @Override
-    public boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type) {
+    public boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type, ReasonType reasonType) {
         assert (type == WorkType.CheckStop || type == WorkType.ForceStop || type == WorkType.Stop);
 
         if (_haDao.hasBeenScheduled(vm.getId(), type)) {
@@ -359,7 +362,7 @@ public boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type) {
             return false;
         }
 
-        HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), type, Step.Scheduled, hostId, vm.getState(), 0, vm.getUpdated());
+        HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), type, Step.Scheduled, hostId, vm.getState(), 0, vm.getUpdated(), reasonType);
         _haDao.persist(work);
         if (logger.isDebugEnabled()) {
             logger.debug("Scheduled " + work);
@@ -368,6 +371,11 @@ public boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type) {
         return true;
     }
 
+    @Override
+    public boolean scheduleStop(VMInstanceVO vm, long hostId, WorkType type) {
+        return scheduleStop(vm, hostId, type, null);
+    }
+
     protected void wakeupWorkers() {
         logger.debug("Wakeup workers HA");
         for (WorkerThread worker : _workers) {
@@ -376,7 +384,7 @@ protected void wakeupWorkers() {
     }
 
     @Override
-    public boolean scheduleMigration(final VMInstanceVO vm) {
+    public boolean scheduleMigration(final VMInstanceVO vm, ReasonType reasonType) {
         if (vm.getHostId() == null) {
             return false;
         }
@@ -390,15 +398,20 @@ public boolean scheduleMigration(final VMInstanceVO vm) {
             return false;
         }
 
-        final HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), WorkType.Migration, Step.Scheduled, vm.getHostId(), vm.getState(), 0, vm.getUpdated());
+        final HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), WorkType.Migration, Step.Scheduled, vm.getHostId(), vm.getState(), 0, vm.getUpdated(), reasonType);
         _haDao.persist(work);
         logger.info("Scheduled migration work of VM {} from host {} with HAWork {}", vm, _hostDao.findById(vm.getHostId()), work);
         wakeupWorkers();
         return true;
     }
 
     @Override
-    public void scheduleRestart(VMInstanceVO vm, boolean investigate) {
+    public boolean scheduleMigration(final VMInstanceVO vm) {
+        return scheduleMigration(vm, null);
+    }
+
+    @Override
+    public void scheduleRestart(VMInstanceVO vm, boolean investigate, ReasonType reasonType) {
         if (!VmHaEnabled.valueIn(vm.getDataCenterId())) {
             String message = String.format("Unable to schedule restart for the VM %s (%d), VM high availability manager is disabled.", vm.getName(), vm.getId());
             if (logger.isDebugEnabled()) {
@@ -490,7 +503,7 @@ public void scheduleRestart(VMInstanceVO vm, boolean investigate) {
         }
 
         HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), WorkType.HA, investigate ? Step.Investigating : Step.Scheduled,
-                hostId != null ? hostId : 0L, vm.getState(), timesTried, vm.getUpdated());
+                hostId != null ? hostId : 0L, vm.getState(), timesTried, vm.getUpdated(), reasonType);
         _haDao.persist(work);
 
         if (logger.isInfoEnabled()) {
@@ -500,6 +513,11 @@ public void scheduleRestart(VMInstanceVO vm, boolean investigate) {
         wakeupWorkers();
     }
 
+    @Override
+    public void scheduleRestart(VMInstanceVO vm, boolean investigate) {
+        scheduleRestart(vm, investigate, null);
+    }
+
     private void startVm(VirtualMachine vm, Map<VirtualMachineProfile.Param, Object> params,
            DeploymentPlanner planner) throws InsufficientCapacityException, ResourceUnavailableException,
             ConcurrentOperationException, OperationTimedoutException {
@@ -561,6 +579,9 @@ protected Long restart(final HaWorkVO work) {
             logger.info("Unable to find vm: " + vmId);
             return null;
         }
+        if (checkAndCancelWorkIfNeeded(work)) {
+            return null;
+        }
 
         logger.info("HA on " + vm);
         if (vm.getState() != work.getPreviousState() || vm.getUpdated() != work.getUpdateTime()) {
@@ -762,6 +783,22 @@ protected Long restart(final HaWorkVO work) {
         return (System.currentTimeMillis() >> 10) + _restartRetryInterval;
     }
 
+    protected boolean checkAndCancelWorkIfNeeded(final HaWorkVO work) {
+        if (!Step.Investigating.equals(work.getStep())) {
+            return false;
+        }
+        if (!CancellableWorkReasonTypes.contains(work.getReasonType())) {
+            return false;
+        }
+        Status hostStatus = investigate(work.getHostId());
+        if (!Status.Up.equals(hostStatus)) {
+            return false;
+        }
+        logger.debug("Cancelling {} as it is not needed anymore", () -> work);
+        work.setStep(Step.Cancelled);
+        return true;
+    }
+
     public Long migrate(final HaWorkVO work) {
         long vmId = work.getInstanceId();
         long srcHostId = work.getHostId();
@@ -772,6 +809,9 @@ public Long migrate(final HaWorkVO work) {
             logger.info("Unable to find vm: " + vmId + ", skipping migrate.");
             return null;
         }
+        if (checkAndCancelWorkIfNeeded(work)) {
+            return null;
+        }
         logger.info("Migration attempt: for VM {}from host {}. Starting attempt: {}/{} times.", vm, srcHost, 1 + work.getTimesTried(), _maxRetries);
         try {
             work.setStep(Step.Migrating);
@@ -791,7 +831,7 @@ public Long migrate(final HaWorkVO work) {
     }
 
     @Override
-    public boolean scheduleDestroy(VMInstanceVO vm, long hostId) {
+    public boolean scheduleDestroy(VMInstanceVO vm, long hostId, ReasonType reasonType) {
         if (!VmHaEnabled.valueIn(vm.getDataCenterId())) {
             String message = String.format("Unable to schedule destroy for the VM %s (%d) on host %d, VM high availability manager is disabled.", vm.getName(), vm.getId(), hostId);
             if (logger.isDebugEnabled()) {
@@ -801,7 +841,7 @@ public boolean scheduleDestroy(VMInstanceVO vm, long hostId) {
             return false;
         }
 
-        final HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), WorkType.Destroy, Step.Scheduled, hostId, vm.getState(), 0, vm.getUpdated());
+        final HaWorkVO work = new HaWorkVO(vm.getId(), vm.getType(), WorkType.Destroy, Step.Scheduled, hostId, vm.getState(), 0, vm.getUpdated(), reasonType);
         _haDao.persist(work);
         if (logger.isDebugEnabled()) {
             logger.debug("Scheduled " + work.toString());
@@ -838,6 +878,9 @@ protected Long destroyVM(final HaWorkVO work) {
             logger.info("No longer can find VM " + work.getInstanceId() + ". Throwing away " + work);
             return null;
         }
+        if (checkAndCancelWorkIfNeeded(work)) {
+            return null;
+        }
         boolean expunge = VirtualMachine.Type.SecondaryStorageVm.equals(vm.getType())
                 || VirtualMachine.Type.ConsoleProxy.equals(vm.getType());
         if (!expunge && VirtualMachine.State.Destroyed.equals(work.getPreviousState())) {
@@ -872,6 +915,9 @@ protected Long stopVM(final HaWorkVO work) throws ConcurrentOperationException {
             work.setStep(Step.Done);
             return null;
         }
+        if (checkAndCancelWorkIfNeeded(work)) {
+            return null;
+        }
         logger.info("Stopping " + vm);
         try {
             if (work.getWorkType() == WorkType.Stop) {
@@ -1057,6 +1103,8 @@ public boolean configure(final String name, final Map<String, Object> xmlParams)
     public boolean start() {
         _stopped = false;
 
+        _haDao.markPendingWorksAsInvestigating();
+
         for (final WorkerThread thread : _workers) {
             thread.start();
         }
@@ -1074,6 +1122,8 @@ public boolean stop() {
 
         _executor.shutdown();
 
+        _haDao.markServerPendingWorksAsInvestigating(_msServer.getId());
+
         return true;
     }
 

server/src/main/java/com/cloud/ha/dao/HighAvailabilityDao.java

@@ -86,4 +86,6 @@ public interface HighAvailabilityDao extends GenericDao<HaWorkVO, Long> {
 
     List<HaWorkVO> listPendingMigrationsForVm(long vmId);
     int expungeByVmList(List<Long> vmIds, Long batchSize);
+    void markPendingWorksAsInvestigating();
+    void markServerPendingWorksAsInvestigating(long managementServerId);
 }