HBASE-23647: Make MasterRegistry the default impl.

Author: bharathv

hbase-client/src/main/java/org/apache/hadoop/hbase/client/ConnectionRegistryFactory.java

@@ -36,7 +36,7 @@ private ConnectionRegistryFactory() {
    */
   static ConnectionRegistry getRegistry(Configuration conf) {
     Class<? extends ConnectionRegistry> clazz = conf.getClass(
-        CLIENT_CONNECTION_REGISTRY_IMPL_CONF_KEY, ZKConnectionRegistry.class,
+        CLIENT_CONNECTION_REGISTRY_IMPL_CONF_KEY, MasterRegistry.class,
         ConnectionRegistry.class);
     return ReflectionUtils.newInstance(clazz, conf);
   }

hbase-client/src/main/java/org/apache/hadoop/hbase/client/MasterRegistry.java

@@ -90,6 +90,7 @@ public class MasterRegistry implements ConnectionRegistry {
     } else {
       finalConf = conf;
     }
+    finalConf.set(MASTER_ADDRS_KEY, conf.get(MASTER_ADDRS_KEY, MASTER_ADDRS_DEFAULT));
     rpcTimeoutMs = (int) Math.min(Integer.MAX_VALUE, conf.getLong(HConstants.HBASE_RPC_TIMEOUT_KEY,
         HConstants.DEFAULT_HBASE_RPC_TIMEOUT));
     masterServers = new HashSet<>();
@@ -146,12 +147,13 @@ private <T, R> RpcCallback<T> getRpcCallBack(CompletableFuture<R> future,
       if (rpcResult == null) {
         future.completeExceptionally(
             new MasterRegistryFetchException(masterServers, hrc.getFailed()));
+        return;
       }
       if (!isValidResp.test(rpcResult)) {
         // Rpc returned ok, but result was malformed.
         future.completeExceptionally(new IOException(
             String.format("Invalid result for request %s. Will be retried", debug)));
-
+        return;
       }
       future.complete(transformResult.apply(rpcResult));
     };

hbase-client/src/main/java/org/apache/hadoop/hbase/client/replication/ReplicationPeerConfigUtil.java

@@ -606,7 +606,6 @@ public static Configuration getPeerClusterConfiguration(Configuration conf,
       compound.addStringMap(peerConfig.getConfiguration());
       return compound;
     }
-
     return otherConf;
   }
 }

hbase-client/src/main/java/org/apache/hadoop/hbase/security/SecurityInfo.java

@@ -49,6 +49,8 @@ public class SecurityInfo {
         new SecurityInfo(SecurityConstants.MASTER_KRB_PRINCIPAL, Kind.HBASE_AUTH_TOKEN));
     infos.put(MasterProtos.HbckService.getDescriptor().getName(),
         new SecurityInfo(SecurityConstants.MASTER_KRB_PRINCIPAL, Kind.HBASE_AUTH_TOKEN));
+    infos.put(MasterProtos.ClientMetaService.getDescriptor().getName(),
+        new SecurityInfo(SecurityConstants.MASTER_KRB_PRINCIPAL, Kind.HBASE_AUTH_TOKEN));
     // NOTE: IF ADDING A NEW SERVICE, BE SURE TO UPDATE HBasePolicyProvider ALSO ELSE
     // new Service will not be found when all is Kerberized!!!!
   }

hbase-common/src/main/java/org/apache/hadoop/hbase/HBaseConfiguration.java

@@ -268,14 +268,19 @@ public static Configuration createClusterConf(Configuration baseConf, String clu
    * used to communicate with distant clusters
    * @param conf configuration object to configure
    * @param key string that contains the 3 required configuratins
-   * @throws IOException
    */
   private static void applyClusterKeyToConf(Configuration conf, String key)
-      throws IOException{
+      throws IOException {
     ZKConfig.ZKClusterKey zkClusterKey = ZKConfig.transformClusterKey(key);
     conf.set(HConstants.ZOOKEEPER_QUORUM, zkClusterKey.getQuorumString());
     conf.setInt(HConstants.ZOOKEEPER_CLIENT_PORT, zkClusterKey.getClientPort());
     conf.set(HConstants.ZOOKEEPER_ZNODE_PARENT, zkClusterKey.getZnodeParent());
+    // Without the right registry, the above configs are useless. Also, we don't use setClass()
+    // here because the ConnectionRegistry* classes are not resolvable from this module.
+    // This will be broken if ZkConnectionRegistry class gets renamed or moved. Is there a better
+    // way?
+    conf.set(HConstants.CLIENT_CONNECTION_REGISTRY_IMPL_CONF_KEY,
+        HConstants.ZK_CONNECTION_REGISTRY_CLASS);
   }
 
   /**

hbase-common/src/main/java/org/apache/hadoop/hbase/HConstants.java

@@ -183,6 +183,10 @@ public enum OperationStatusCode {
 
   public static final String MASTER_ADDRS_DEFAULT =  "localhost:" + DEFAULT_MASTER_PORT;
 
+  /** Full class name of the Zookeeper based connection registry implementation */
+  public static final String ZK_CONNECTION_REGISTRY_CLASS =
+      "org.apache.hadoop.hbase.client.ZKConnectionRegistry";
+
   /** Configuration to enable hedged reads on master registry **/
   public static final String MASTER_REGISTRY_ENABLE_HEDGED_READS_KEY =
       "hbase.client.master_registry.enable_hedged_reads";

hbase-server/src/main/java/org/apache/hadoop/hbase/LocalHBaseCluster.java

@@ -91,7 +91,7 @@ public LocalHBaseCluster(final Configuration conf)
    */
   public LocalHBaseCluster(final Configuration conf, final int noRegionServers)
   throws IOException {
-    this(conf, 1, noRegionServers, getMasterImplementation(conf),
+    this(conf, 1, 0, noRegionServers, getMasterImplementation(conf),
         getRegionServerImplementation(conf));
   }
 
@@ -106,7 +106,7 @@ public LocalHBaseCluster(final Configuration conf, final int noRegionServers)
   public LocalHBaseCluster(final Configuration conf, final int noMasters,
       final int noRegionServers)
   throws IOException {
-    this(conf, noMasters, noRegionServers, getMasterImplementation(conf),
+    this(conf, noMasters, 0, noRegionServers, getMasterImplementation(conf),
         getRegionServerImplementation(conf));
   }
 
@@ -122,6 +122,12 @@ private static Class<? extends HMaster> getMasterImplementation(final Configurat
        HMaster.class);
   }
 
+  public LocalHBaseCluster(final Configuration conf, final int noMasters, final int noRegionServers,
+      final Class<? extends HMaster> masterClass,
+      final Class<? extends HRegionServer> regionServerClass) throws IOException {
+    this(conf, noMasters, 0, noRegionServers, masterClass, regionServerClass);
+  }
+
   /**
    * Constructor.
    * @param conf Configuration to use.  Post construction has the master's
@@ -134,9 +140,9 @@ private static Class<? extends HMaster> getMasterImplementation(final Configurat
    */
   @SuppressWarnings("unchecked")
   public LocalHBaseCluster(final Configuration conf, final int noMasters,
-    final int noRegionServers, final Class<? extends HMaster> masterClass,
-    final Class<? extends HRegionServer> regionServerClass)
-  throws IOException {
+      final int noAlwaysStandByMasters, final int noRegionServers,
+      final Class<? extends HMaster> masterClass,
+      final Class<? extends HRegionServer> regionServerClass) throws IOException {
     this.conf = conf;
 
     // When active, if a port selection is default then we switch to random
@@ -170,24 +176,22 @@ public LocalHBaseCluster(final Configuration conf, final int noMasters,
     this.masterClass = (Class<? extends HMaster>)
       conf.getClass(HConstants.MASTER_IMPL, masterClass);
     // Start the HMasters.
-    for (int i = 0; i < noMasters; i++) {
+    int i;
+    for (i = 0; i < noMasters; i++) {
       addMaster(new Configuration(conf), i);
     }
-
-    // Populate the master address host ports in the config. This is needed if a master based
-    // registry is configured for client metadata services (HBASE-18095)
-    List<String> masterHostPorts = new ArrayList<>();
-    getMasters().forEach(masterThread ->
-        masterHostPorts.add(masterThread.getMaster().getServerName().getAddress().toString()));
-    conf.set(HConstants.MASTER_ADDRS_KEY, String.join(",", masterHostPorts));
-
+    for (int j = 0; j < noAlwaysStandByMasters; j++) {
+      Configuration c = new Configuration(conf);
+      c.set(HConstants.MASTER_IMPL, "org.apache.hadoop.hbase.master.AlwaysStandByHMaster");
+      addMaster(c, i + j);
+    }
     // Start the HRegionServers.
     this.regionServerClass =
       (Class<? extends HRegionServer>)conf.getClass(HConstants.REGION_SERVER_IMPL,
        regionServerClass);
 
-    for (int i = 0; i < noRegionServers; i++) {
-      addRegionServer(new Configuration(conf), i);
+    for (int j = 0; j < noRegionServers; j++) {
+      addRegionServer(new Configuration(conf), j);
     }
   }
 
@@ -233,8 +237,13 @@ public JVMClusterUtil.MasterThread addMaster(Configuration c, final int index)
     // its Connection instance rather than share (see HBASE_INSTANCES down in
     // the guts of ConnectionManager.
     JVMClusterUtil.MasterThread mt = JVMClusterUtil.createMasterThread(c,
-        (Class<? extends HMaster>) conf.getClass(HConstants.MASTER_IMPL, this.masterClass), index);
+        (Class<? extends HMaster>) c.getClass(HConstants.MASTER_IMPL, this.masterClass), index);
     this.masterThreads.add(mt);
+    // Refresh the master address config.
+    List<String> masterHostPorts = new ArrayList<>();
+    getMasters().forEach(masterThread ->
+        masterHostPorts.add(masterThread.getMaster().getServerName().getAddress().toString()));
+    conf.set(HConstants.MASTER_ADDRS_KEY, String.join(",", masterHostPorts));
     return mt;
   }
 

hbase-server/src/main/java/org/apache/hadoop/hbase/master/ActiveMasterManager.java

@@ -56,15 +56,15 @@ public class ActiveMasterManager extends ZKListener {
   final AtomicBoolean clusterHasActiveMaster = new AtomicBoolean(false);
   final AtomicBoolean clusterShutDown = new AtomicBoolean(false);
 
-  // This server's information.
-  private final ServerName sn;
-  private int infoPort;
-  private final Server master;
+  // This server's information. Package-private for child implementations.
+  int infoPort;
+  final ServerName sn;
+  final Server master;
 
   // Active master's server name. Invalidated anytime active master changes (based on ZK
   // notifications) and lazily fetched on-demand.
   // ServerName is immutable, so we don't need heavy synchronization around it.
-  private volatile ServerName activeMasterServerName;
+  volatile ServerName activeMasterServerName;
 
   /**
    * @param watcher ZK watcher

hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java

@@ -575,7 +575,7 @@ public HMaster(final Configuration conf) throws IOException {
       // Some unit tests don't need a cluster, so no zookeeper at all
       if (!conf.getBoolean("hbase.testing.nocluster", false)) {
         this.metaRegionLocationCache = new MetaRegionLocationCache(this.zooKeeper);
-        this.activeMasterManager = new ActiveMasterManager(zooKeeper, this.serverName, this);
+        this.activeMasterManager = createActiveMasterManager(zooKeeper, serverName, this);
       } else {
         this.metaRegionLocationCache = null;
         this.activeMasterManager = null;
@@ -589,6 +589,15 @@ public HMaster(final Configuration conf) throws IOException {
     }
   }
 
+  /**
+   * Protected to have custom implementations in tests override the default ActiveMaster
+   * implementation.
+   */
+  protected ActiveMasterManager createActiveMasterManager(
+      ZKWatcher zk, ServerName sn, org.apache.hadoop.hbase.Server server) {
+    return new ActiveMasterManager(zk, sn, server);
+  }
+
   @Override
   protected String getUseThisHostnameInstead(Configuration conf) {
     return conf.get(MASTER_HOSTNAME_KEY);

hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java

@@ -789,8 +789,17 @@ public boolean registerService(com.google.protobuf.Service instance) {
     return true;
   }
 
-  private Configuration unsetClientZookeeperQuorum() {
+  private Configuration cleanupConfiguration() {
     Configuration conf = this.conf;
+    // We use ZKConnectionRegistry for all the internal communication, primarily for these reasons:
+    // - Decouples RS and master life cycles. RegionServers can continue be up independent of
+    //   masters' availability.
+    // - Configuration management for region servers (cluster internal) is much simpler when adding
+    //   new masters or removing existing masters, since only clients' config needs to be updated.
+    // - We need to retain ZKConnectionRegistry for replication use anyway, so we just extend it for
+    //   other internal connections too.
+    conf.set(HConstants.CLIENT_CONNECTION_REGISTRY_IMPL_CONF_KEY,
+        HConstants.ZK_CONNECTION_REGISTRY_CLASS);
     if (conf.get(HConstants.CLIENT_ZOOKEEPER_QUORUM) != null) {
       // Use server ZK cluster for server-issued connections, so we clone
       // the conf and unset the client ZK related properties
@@ -824,7 +833,7 @@ public String getClusterId() {
    */
   protected final synchronized void setupClusterConnection() throws IOException {
     if (asyncClusterConnection == null) {
-      Configuration conf = unsetClientZookeeperQuorum();
+      Configuration conf = cleanupConfiguration();
       InetSocketAddress localAddress = new InetSocketAddress(this.rpcServices.isa.getAddress(), 0);
       User user = userProvider.getCurrent();
       asyncClusterConnection =