HBASE-23281: Track meta region locations in masters (apache#830)

* HBASE-23281: Track meta region changes on masters

This patch adds a simple cache that tracks the meta region replica
locations. It keeps an eye on the region movements so that the
cached locations are not stale.

This information is used for servicing client RPCs for connections
that use master based registry (HBASE-18095). The RPC end points
will be added in a separate patch.

Signed-off-by: Nick Dimiduk <ndimiduk@apache.org>

Author: bharathv

hbase-client/src/main/java/org/apache/hadoop/hbase/shaded/protobuf/ProtobufUtil.java

@@ -1,4 +1,4 @@
-/**
+/*
  * Licensed to the Apache Software Foundation (ASF) under one
  * or more contributor license agreements.  See the NOTICE file
  * distributed with this work for additional information
@@ -80,6 +80,7 @@
 import org.apache.hadoop.hbase.client.Put;
 import org.apache.hadoop.hbase.client.RegionInfoBuilder;
 import org.apache.hadoop.hbase.client.RegionLoadStats;
+import org.apache.hadoop.hbase.client.RegionReplicaUtil;
 import org.apache.hadoop.hbase.client.RegionStatesCount;
 import org.apache.hadoop.hbase.client.Result;
 import org.apache.hadoop.hbase.client.Scan;
@@ -93,6 +94,7 @@
 import org.apache.hadoop.hbase.filter.ByteArrayComparable;
 import org.apache.hadoop.hbase.filter.Filter;
 import org.apache.hadoop.hbase.io.TimeRange;
+import org.apache.hadoop.hbase.master.RegionState;
 import org.apache.hadoop.hbase.protobuf.ProtobufMagic;
 import org.apache.hadoop.hbase.protobuf.ProtobufMessageConverter;
 import org.apache.hadoop.hbase.quotas.QuotaScope;
@@ -3067,6 +3069,44 @@ public static ProcedureDescription buildProcedureDescription(String signature, S
     return builder.build();
   }
 
+  /**
+   * Get the Meta region state from the passed data bytes. Can handle both old and new style
+   * server names.
+   * @param data protobuf serialized data with meta server name.
+   * @param replicaId replica ID for this region
+   * @return RegionState instance corresponding to the serialized data.
+   * @throws DeserializationException if the data is invalid.
+   */
+  public static RegionState parseMetaRegionStateFrom(final byte[] data, int replicaId)
+      throws DeserializationException {
+    RegionState.State state = RegionState.State.OPEN;
+    ServerName serverName;
+    if (data != null && data.length > 0 && ProtobufUtil.isPBMagicPrefix(data)) {
+      try {
+        int prefixLen = ProtobufUtil.lengthOfPBMagic();
+        ZooKeeperProtos.MetaRegionServer rl =
+            ZooKeeperProtos.MetaRegionServer.parser().parseFrom(data, prefixLen,
+                data.length - prefixLen);
+        if (rl.hasState()) {
+          state = RegionState.State.convert(rl.getState());
+        }
+        HBaseProtos.ServerName sn = rl.getServer();
+        serverName = ServerName.valueOf(
+            sn.getHostName(), sn.getPort(), sn.getStartCode());
+      } catch (InvalidProtocolBufferException e) {
+        throw new DeserializationException("Unable to parse meta region location");
+      }
+    } else {
+      // old style of meta region location?
+      serverName = parseServerNameFrom(data);
+    }
+    if (serverName == null) {
+      state = RegionState.State.OFFLINE;
+    }
+    return new RegionState(RegionReplicaUtil.getRegionInfoForReplica(
+        RegionInfoBuilder.FIRST_META_REGIONINFO, replicaId), state, serverName);
+  }
+
   /**
    * Get a ServerName from the passed in data bytes.
    * @param data Data with a serialize server name in it; can handle the old style

hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZNodePaths.java

@@ -41,7 +41,8 @@ public class ZNodePaths {
   // TODO: Replace this with ZooKeeper constant when ZOOKEEPER-277 is resolved.
   public static final char ZNODE_PATH_SEPARATOR = '/';
 
-  private static final String META_ZNODE_PREFIX = "meta-region-server";
+  public static final String META_ZNODE_PREFIX_CONF_KEY = "zookeeper.znode.metaserver";
+  public static final String META_ZNODE_PREFIX = "meta-region-server";
   private static final String DEFAULT_SNAPSHOT_CLEANUP_ZNODE = "snapshot-cleanup";
 
   // base znode for this cluster
@@ -104,7 +105,7 @@ public class ZNodePaths {
   public ZNodePaths(Configuration conf) {
     baseZNode = conf.get(ZOOKEEPER_ZNODE_PARENT, DEFAULT_ZOOKEEPER_ZNODE_PARENT);
     ImmutableMap.Builder<Integer, String> builder = ImmutableMap.builder();
-    metaZNodePrefix = conf.get("zookeeper.znode.metaserver", META_ZNODE_PREFIX);
+    metaZNodePrefix = conf.get(META_ZNODE_PREFIX_CONF_KEY, META_ZNODE_PREFIX);
     String defaultMetaReplicaZNode = ZNodePaths.joinZNode(baseZNode, metaZNodePrefix);
     builder.put(DEFAULT_REPLICA_ID, defaultMetaReplicaZNode);
     int numMetaReplicas = conf.getInt(META_REPLICAS_NUM, DEFAULT_META_REPLICA_NUM);
@@ -189,7 +190,19 @@ public String getZNodeForReplica(int replicaId) {
   }
 
   /**
-   * Parse the meta replicaId from the passed znode name.
+   * Parses the meta replicaId from the passed path.
+   * @param path the name of the full path which includes baseZNode.
+   * @return replicaId
+   */
+  public int getMetaReplicaIdFromPath(String path) {
+    // Extract the znode from path. The prefix is of the following format.
+    // baseZNode + PATH_SEPARATOR.
+    int prefixLen = baseZNode.length() + 1;
+    return getMetaReplicaIdFromZnode(path.substring(prefixLen));
+  }
+
+  /**
+   * Parse the meta replicaId from the passed znode
    * @param znode the name of the znode, does not include baseZNode
    * @return replicaId
    */

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

@@ -1,4 +1,4 @@
-/**
+/*
  * Licensed to the Apache Software Foundation (ASF) under one
  * or more contributor license agreements.  See the NOTICE file
  * distributed with this work for additional information
@@ -356,6 +356,12 @@ public void run() {
   // manager of assignment nodes in zookeeper
   private AssignmentManager assignmentManager;
 
+  /**
+   * Cache for the meta region replica's locations. Also tracks their changes to avoid stale
+   * cache entries.
+   */
+  private final MetaRegionLocationCache metaRegionLocationCache;
+
   // manager of replication
   private ReplicationPeerManager replicationPeerManager;
 
@@ -508,8 +514,7 @@ public void doGet(HttpServletRequest request,
    * {@link #finishActiveMasterInitialization(MonitoredTask)} after the master becomes the
    * active one.
    */
-  public HMaster(final Configuration conf)
-      throws IOException, KeeperException {
+  public HMaster(final Configuration conf) throws IOException {
     super(conf);
     TraceUtil.initTracer(conf);
     try {
@@ -522,7 +527,6 @@ public HMaster(final Configuration conf)
       } else {
         maintenanceMode = false;
       }
-
       this.rsFatals = new MemoryBoundedLogMessageBuffer(
           conf.getLong("hbase.master.buffer.for.rs.fatals", 1 * 1024 * 1024));
       LOG.info("hbase.rootdir={}, hbase.cluster.distributed={}", getDataRootDir(),
@@ -570,8 +574,10 @@ public HMaster(final Configuration conf)
 
       // 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);
       } else {
+        this.metaRegionLocationCache = null;
         this.activeMasterManager = null;
       }
       cachedClusterId = new CachedClusterId(conf);
@@ -3838,4 +3844,8 @@ public void runReplicationBarrierCleaner() {
       rbc.chore();
     }
   }
+
+  public MetaRegionLocationCache getMetaRegionLocationCache() {
+    return this.metaRegionLocationCache;
+  }
 }