HDFS-17769. Allows client to actively retry to Active NameNode when the Observer NameNode is too far behind client state id.

Author: guoyz50

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java

@@ -273,10 +273,6 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
       HdfsClientConfigKeys.DeprecatedKeys.DFS_NAMENODE_REDUNDANCY_CONSIDERLOAD_KEY;
   public static final boolean DFS_NAMENODE_REDUNDANCY_CONSIDERLOAD_DEFAULT =
       true;
-  public static final String DFS_NAMENODE_OBSERVER_TOO_STALE_RETRY_ACTIVE_ENABLE  =
-          "dfs.namenode.observer.too.stale.retry.active.enable";
-  public static final boolean DFS_NAMENODE_OBSERVER_TOO_STALE_RETRY_ACTIVE_ENABLE_DEFAULT  =
-          false;
   public static final String
       DFS_NAMENODE_REDUNDANCY_CONSIDERLOADBYSTORAGETYPE_KEY =
       "dfs.namenode.redundancy.considerLoadByStorageType";

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java

@@ -694,8 +694,6 @@ private boolean isFromProxyUser(CallerContext ctx) {
 
   private final MessageDigest digest;
 
-  private final boolean isRetryActive;
-
   /**
    * Notify that loading of this FSDirectory is complete, and
    * it is imageLoaded for use
@@ -1080,9 +1078,6 @@ static FSNamesystem loadFromDisk(Configuration conf) throws IOException {
       this.isGetBlocksCheckOperationEnabled = conf.getBoolean(
           DFSConfigKeys.DFS_NAMENODE_GETBLOCKS_CHECK_OPERATION_KEY,
           DFSConfigKeys.DFS_NAMENODE_GETBLOCKS_CHECK_OPERATION_DEFAULT);
-      this.isRetryActive = conf.getBoolean(
-              DFSConfigKeys.DFS_NAMENODE_OBSERVER_TOO_STALE_RETRY_ACTIVE_ENABLE,
-              DFSConfigKeys.DFS_NAMENODE_OBSERVER_TOO_STALE_RETRY_ACTIVE_ENABLE_DEFAULT);
 
     } catch(IOException e) {
       LOG.error(getClass().getSimpleName() + " initialization failed.", e);
@@ -1143,10 +1138,6 @@ boolean isSnapshotTrashRootEnabled() {
     return isSnapshotTrashRootEnabled;
   }
 
-  public boolean isRetryActive() {
-    return isRetryActive;
-  }
-
   void lockRetryCache() {
     if (retryCache != null) {
       retryCache.lock();

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/GlobalStateIdContext.java

@@ -26,7 +26,6 @@
 import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.classification.InterfaceStability;
 import org.apache.hadoop.ha.HAServiceProtocol.HAServiceState;
-import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.protocol.ClientProtocol;
 import org.apache.hadoop.hdfs.server.namenode.ha.ObserverReadProxyProvider;
 import org.apache.hadoop.hdfs.server.namenode.ha.ReadOnly;
@@ -158,18 +157,9 @@ public long receiveRequestState(RpcRequestHeaderProto header,
         ESTIMATED_TRANSACTIONS_PER_SECOND
             * TimeUnit.MILLISECONDS.toSeconds(clientWaitTime)
             * ESTIMATED_SERVER_TIME_MULTIPLIER) {
-      if (namesystem.isRetryActive()) {
-        String message = "Retrying to Active NameNode, Observer Node is too far behind: serverStateId = " + serverStateId + " clientStateId = " + clientStateId;
-        FSNamesystem.LOG.warn(message);
-        throw new ObserverRetryOnActiveException(message);
-      } else {
-        throw new RetriableException(
-                String.format(
-                        "%s is not enabled, Continue Retrying to Observer NameNode, Observer Node is too far behind: serverStateId = %d clientStateId = %d",
-                        DFSConfigKeys.DFS_NAMENODE_OBSERVER_TOO_STALE_RETRY_ACTIVE_ENABLE,
-                        serverStateId, clientStateId
-                ));
-      }
+      throw new ObserverRetryOnActiveException("Retrying to Active NameNode, Observer Node is too"
+          + " far behind: serverStateId = " + serverStateId
+          + " clientStateId = " + clientStateId);
     }
     return clientStateId;
   }

hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml

@@ -3753,13 +3753,6 @@
   </description>
 </property>
 
-<property>
-  <name>dfs.namenode.observer.too.stale.retry.active.enable</name>
-  <value>false</value>
-  <description>When enabled, allows client to actively retry to Active NameNode when the Observer NameNode is too far behind client state id.
-  </description>
-</property>
-
 <property>
   <name>dfs.namenode.top.windows.minutes</name>
   <value>1,5,25</value>

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/ha/TestObserverNode.java

@@ -53,7 +53,6 @@
 import org.apache.hadoop.fs.permission.FsPermission;
 import org.apache.hadoop.ha.HAServiceProtocol.HAServiceState;
 import org.apache.hadoop.ha.ServiceFailedException;
-import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
@@ -107,7 +106,6 @@ public static void startUpCluster() throws Exception {
     // Observer and immediately try to read from it.
     conf.setTimeDuration(
         OBSERVER_PROBE_RETRY_PERIOD_KEY, 0, TimeUnit.MILLISECONDS);
-    conf.setBoolean(DFSConfigKeys.DFS_NAMENODE_OBSERVER_TOO_STALE_RETRY_ACTIVE_ENABLE, true);
     qjmhaCluster = HATestUtil.setUpObserverCluster(conf, 1, 1, true);
     dfsCluster = qjmhaCluster.getDfsCluster();
   }