ARTEMIS-4325 - Ability to failback after failover

Author: AntonRoskvist

artemis-core-client/src/main/java/org/apache/activemq/artemis/api/config/ServerLocatorConfig.java

@@ -45,6 +45,7 @@ public class ServerLocatorConfig {
    public int reconnectAttempts = ActiveMQClient.DEFAULT_RECONNECT_ATTEMPTS;
    public int initialConnectAttempts = ActiveMQClient.INITIAL_CONNECT_ATTEMPTS;
    public int failoverAttempts = ActiveMQClient.DEFAULT_FAILOVER_ATTEMPTS;
+   public int failbackAttempts = ActiveMQClient.DEFAULT_FAILBACK_ATTEMPTS;
    public int initialMessagePacketSize = ActiveMQClient.DEFAULT_INITIAL_MESSAGE_PACKET_SIZE;
    public boolean cacheLargeMessagesClient = ActiveMQClient.DEFAULT_CACHE_LARGE_MESSAGE_CLIENT;
    public int compressionLevel = ActiveMQClient.DEFAULT_COMPRESSION_LEVEL;
@@ -83,6 +84,7 @@ public ServerLocatorConfig(final ServerLocatorConfig locator) {
       reconnectAttempts = locator.reconnectAttempts;
       initialConnectAttempts = locator.initialConnectAttempts;
       failoverAttempts = locator.failoverAttempts;
+      failbackAttempts = locator.failbackAttempts;
       initialMessagePacketSize = locator.initialMessagePacketSize;
       useTopologyForLoadBalancing = locator.useTopologyForLoadBalancing;
       compressionLevel = locator.compressionLevel;

artemis-core-client/src/main/java/org/apache/activemq/artemis/api/core/client/ActiveMQClient.java

@@ -117,6 +117,8 @@ public final class ActiveMQClient {
 
    public static final int DEFAULT_FAILOVER_ATTEMPTS = 0;
 
+   public static final int DEFAULT_FAILBACK_ATTEMPTS = 0;
+
    @Deprecated
    public static final boolean DEFAULT_FAILOVER_ON_INITIAL_CONNECTION = false;
 

artemis-core-client/src/main/java/org/apache/activemq/artemis/api/core/client/ServerLocator.java

@@ -668,6 +668,21 @@ ClientSessionFactory createSessionFactory(TransportConfiguration transportConfig
     */
    int getFailoverAttempts();
 
+   /**
+    * Sets the maximum number of failback attempts to establish a new conection to the original broker after a failover.
+    * <p>
+    * Value must be -1 (to try infinitely), 0 (to never atempt failback) or greater than 0.
+    *
+    * @param attempts maximum number of failback attempts after a successful failover
+    * @return this ServerLocator
+    */
+   ServerLocator setFailbackAttempts(int attempts);
+
+   /**
+    * @return the number of failback attempts after a successful failover
+    */
+   int getFailbackAttempts();
+
    /**
     * Returns true if the client will automatically attempt to connect to the backup server if the initial
     * connection to the live server fails

artemis-core-client/src/main/java/org/apache/activemq/artemis/core/client/impl/ClientSessionFactoryImpl.java

@@ -77,6 +77,8 @@
 import java.lang.invoke.MethodHandles;
 import java.util.function.BiPredicate;
 
+import static org.apache.activemq.artemis.api.core.ActiveMQExceptionType.DISCONNECTED;
+
 public class ClientSessionFactoryImpl implements ClientSessionFactoryInternal, ClientConnectionLifeCycleListener {
 
    private static final Logger logger = LoggerFactory.getLogger(MethodHandles.lookup().lookupClass());
@@ -93,6 +95,8 @@ public class ClientSessionFactoryImpl implements ClientSessionFactoryInternal, C
 
    private volatile TransportConfiguration backupConnectorConfig;
 
+   private TransportConfiguration failbackConnectorConfig;
+
    private ConnectorFactory connectorFactory;
 
    private final long callTimeout;
@@ -135,6 +139,8 @@ public class ClientSessionFactoryImpl implements ClientSessionFactoryInternal, C
 
    private int failoverAttempts;
 
+   private int failbackAttempts;
+
    private final Set<SessionFailureListener> listeners = new ConcurrentHashSet<>();
 
    private final Set<FailoverEventListener> failoverListeners = new ConcurrentHashSet<>();
@@ -144,6 +150,8 @@ public class ClientSessionFactoryImpl implements ClientSessionFactoryInternal, C
    private Future<?> pingerFuture;
    private PingRunnable pingRunnable;
 
+   private FailbackRunnable failbackRunnable;
+
    private final List<Interceptor> incomingInterceptors;
 
    private final List<Interceptor> outgoingInterceptors;
@@ -244,6 +252,8 @@ public ClientSessionFactoryImpl(final ServerLocatorInternal serverLocator,
 
       this.failoverAttempts = locatorConfig.failoverAttempts;
 
+      this.failbackAttempts = locatorConfig.failbackAttempts;
+
       this.scheduledThreadPool = scheduledThreadPool;
 
       this.threadPool = threadPool;
@@ -722,6 +732,12 @@ private void failoverOrReconnect(final Object connectionID,
                int connectorsCount = 0;
                int failoverRetries = 0;
                long failoverRetryInterval = retryInterval;
+
+               //Save current connector config for failback purposes
+               if (failbackAttempts != 0 && failbackConnectorConfig == null) {
+                  failbackConnectorConfig = connectorConfig;
+               }
+
                Pair<TransportConfiguration, TransportConfiguration> connectorPair;
                BiPredicate<Boolean, Integer> failoverRetryPredicate =
                   (reconnected, retries) -> clientProtocolManager.isAlive() &&
@@ -815,6 +831,128 @@ private void failoverOrReconnect(final Object connectionID,
       }
    }
 
+   private void failback(final ActiveMQException me,
+                         final TransportConfiguration previousConnectorConfig) {
+
+      logger.debug("Original node has come back online, performing failback now");
+
+      for (ClientSessionInternal session : sessions) {
+         SessionContext context = session.getSessionContext();
+         if (context instanceof ActiveMQSessionContext) {
+            ActiveMQSessionContext sessionContext = (ActiveMQSessionContext) context;
+            if (sessionContext.isKilled()) {
+               setReconnectAttempts(0);
+            }
+         }
+      }
+
+      Set<ClientSessionInternal> sessionsToClose = null;
+      if (!clientProtocolManager.isAlive()) {
+         return;
+      }
+
+      Lock localFailoverLock = lockFailover();
+
+      try {
+
+         callFailoverListeners(FailoverEventType.FAILURE_DETECTED);
+         callSessionFailureListeners(me, false, false, null);
+
+         if (clientProtocolManager.cleanupBeforeFailover(me)) {
+
+            RemotingConnection oldConnection = connection;
+
+            connection = null;
+
+            Connector localConnector = connector;
+            if (localConnector != null) {
+               try {
+                  localConnector.close();
+               } catch (Exception ignore) {
+                  // no-op
+               }
+            }
+
+            cancelScheduledTasks();
+
+            connector = null;
+
+            HashSet<ClientSessionInternal> sessionsToFailover;
+            synchronized (sessions) {
+               sessionsToFailover = new HashSet<>(sessions);
+            }
+
+            // Notify sessions before failover.
+            for (ClientSessionInternal session : sessionsToFailover) {
+               session.preHandleFailover(connection);
+            }
+
+            boolean sessionsReconnected = false;
+
+            connectorConfig = previousConnectorConfig;
+            currentConnectorConfig = previousConnectorConfig;
+
+            getConnection();
+
+            if (connection != null) {
+               sessionsReconnected = reconnectSessions(sessionsToFailover, oldConnection, me);
+
+               if (!sessionsReconnected) {
+                  if (oldConnection != null) {
+                     oldConnection.destroy();
+                  }
+
+                  oldConnection = connection;
+                  connection = null;
+               }
+            }
+
+            // Notify sessions after failover.
+            for (ClientSessionInternal session : sessionsToFailover) {
+               session.postHandleFailover(connection, sessionsReconnected);
+            }
+
+            if (oldConnection != null) {
+               oldConnection.destroy();
+            }
+
+            if (connection != null) {
+               callFailoverListeners(FailoverEventType.FAILOVER_COMPLETED);
+
+            }
+         }
+
+         if (connection == null) {
+            synchronized (sessions) {
+               sessionsToClose = new HashSet<>(sessions);
+            }
+            callFailoverListeners(FailoverEventType.FAILOVER_FAILED);
+            callSessionFailureListeners(me, true, false, null);
+         }
+      } finally {
+         localFailoverLock.unlock();
+      }
+
+      // This needs to be outside the failover lock to prevent deadlock
+      if (connection != null) {
+         callSessionFailureListeners(me, true, true);
+      }
+
+      if (sessionsToClose != null) {
+         // If connection is null it means we didn't succeed in failing over or reconnecting
+         // so we close all the sessions, so they will throw exceptions when attempted to be used
+
+         for (ClientSessionInternal session : sessionsToClose) {
+            try {
+               session.cleanUp(true);
+            } catch (Exception cause) {
+               ActiveMQClientLogger.LOGGER.failedToCleanupSession(cause);
+            }
+         }
+      }
+
+   }
+
    private ClientSession createSessionInternal(final String rawUsername,
                                                final String rawPassword,
                                                final boolean xa,
@@ -1018,6 +1156,10 @@ public boolean waitForRetry(long interval) {
       return false;
    }
 
+   private long getRetryInterval() {
+      return retryInterval;
+   }
+
    private void cancelScheduledTasks() {
       Future<?> pingerFutureLocal = pingerFuture;
       if (pingerFutureLocal != null) {
@@ -1027,8 +1169,13 @@ private void cancelScheduledTasks() {
       if (pingRunnableLocal != null) {
          pingRunnableLocal.cancel();
       }
+      FailbackRunnable failbackRunnableLocal = failbackRunnable;
+      if (failbackRunnableLocal != null) {
+         failbackRunnableLocal.cancel();
+      }
       pingerFuture = null;
       pingRunnable = null;
+      failbackRunnable = null;
    }
 
    private void checkCloseConnection() {
@@ -1492,6 +1639,68 @@ public synchronized void cancel() {
       }
    }
 
+   private void attemptFailback() {
+      if (failbackRunnable == null) {
+         failbackRunnable = new FailbackRunnable();
+      }
+      threadPool.execute(failbackRunnable);
+   }
+
+   private class FailbackRunnable implements Runnable {
+      private boolean first = true;
+      private boolean cancelled;
+
+      @Override
+      public synchronized void run() {
+
+         if (!first) {
+            return;
+         }
+
+         first = false;
+
+         logger.debug("Attempting failback. Trying to reach {} for failback", failbackConnectorConfig.toString());
+
+         int attempts = 0;
+         long failbackRetryInterval = getRetryInterval();
+
+         ConnectorFactory transportConnectorFactory;
+         Connector transportConnector;
+         Connection transportConnection;
+
+         while (!cancelled && (failbackAttempts == -1 || attempts++ < failbackAttempts)) {
+
+            waitForRetry(failbackRetryInterval);
+            failbackRetryInterval = getNextRetryInterval(failbackRetryInterval);
+
+            transportConnectorFactory = instantiateConnectorFactory(failbackConnectorConfig.getFactoryClassName());
+            transportConnector = createConnector(transportConnectorFactory, failbackConnectorConfig);
+            transportConnection = openTransportConnection(transportConnector);
+
+            if (transportConnection != null) {
+               transportConnector.close();
+               transportConnection.close();
+               ActiveMQException exception = new ActiveMQException("Failing back to original broker: " + failbackConnectorConfig.toString(), DISCONNECTED);
+               failback(exception, failbackConnectorConfig);
+               break;
+            }
+
+         }
+
+         if (failbackConnectorConfig.equals(currentConnectorConfig)) {
+            failbackConnectorConfig = null;
+         }
+
+         first = true;
+
+      }
+
+      public synchronized void cancel() {
+         cancelled = true;
+      }
+
+   }
+
    protected RemotingConnection establishNewConnection() {
       Connection transportConnection = createTransportConnection();
 
@@ -1572,6 +1781,13 @@ public void notifyNodeUp(long uniqueEventID,
                                boolean isLast) {
 
          try {
+
+            if (failbackConnectorConfig != null && connectorPair.getA() != null && TransportConfigurationUtil.isSameHost(connectorPair.getA(), failbackConnectorConfig)) {
+               if (!currentConnectorConfig.equals(failbackConnectorConfig) && failbackRunnable == null) {
+                  attemptFailback();
+               }
+            }
+
             // if it is our connector then set the live id used for failover
             if (connectorPair.getA() != null && TransportConfigurationUtil.isSameHost(connectorPair.getA(), currentConnectorConfig)) {
                liveNodeID = nodeID;

artemis-core-client/src/main/java/org/apache/activemq/artemis/core/client/impl/ServerLocatorImpl.java

@@ -1199,6 +1199,18 @@ public int getFailoverAttempts() {
       return config.failoverAttempts;
    }
 
+   @Override
+   public ServerLocatorImpl setFailbackAttempts(int attempts) {
+      checkWrite();
+      this.config.failbackAttempts = attempts;
+      return this;
+   }
+
+   @Override
+   public int getFailbackAttempts() {
+      return config.failbackAttempts;
+   }
+
    @Deprecated
    @Override
    public boolean isFailoverOnInitialConnection() {

docs/user-manual/client-failover.adoc

@@ -48,6 +48,16 @@ Set `failoverAttempts` to any non-zero value to reconnect to other live servers,
 
 If `reconnectAttempts` value is not zero then the client will try to reconnect to other live servers only after all attempts to <<reconnect-to-the-same-server,reconnect to the same server>> or <<reconnect-to-the-backup-server,reconnect to the backup server>> fail.
 
+== Failing back after a successful failover
+
+It is also possile to have the client keep trying to reconnect to the original live server after a successful failover. In this case the failover connection will get disconnected and moved back to the original server if it comes back online at some later time.
+
+This is controlled by the property `failbackAttempts`
+
+If set to a non-zero value then following a failover the client will try to reach the original server continiously according to the configured `retryInterval`  until it reaches the number of configured `failbackAttempts`.
+
+Setting this parameter to `-1` means try forever. Default value is `0`
+
 == Session reconnection
 
 When clients <<reconnect-to-the-same-server,reconnect to the same server>> after a restart, <<reconnect-to-the-backup-server,reconnect to the backup server>> or <<reconnect-to-other-live-servers,reconnect to other live servers>> any sessions will no longer exist on the server and it won't be possible to 100% transparently re-attach to them.