HBASE-28638 Fail-fast retry limit for specific errors to recover from remote procedure failure using server crash (#6564) (#6462)

Signed-off-by: Duo Zhang <[email protected]>

Author: virajjasani

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

@@ -159,6 +159,7 @@
 import org.apache.hadoop.hbase.master.procedure.ModifyTableProcedure;
 import org.apache.hadoop.hbase.master.procedure.ProcedurePrepareLatch;
 import org.apache.hadoop.hbase.master.procedure.ProcedureSyncWait;
+import org.apache.hadoop.hbase.master.procedure.RSProcedureDispatcher;
 import org.apache.hadoop.hbase.master.procedure.ReopenTableRegionsProcedure;
 import org.apache.hadoop.hbase.master.procedure.ServerCrashProcedure;
 import org.apache.hadoop.hbase.master.procedure.TruncateTableProcedure;
@@ -231,6 +232,7 @@
 import org.apache.hadoop.hbase.util.IdLock;
 import org.apache.hadoop.hbase.util.ModifyRegionUtils;
 import org.apache.hadoop.hbase.util.Pair;
+import org.apache.hadoop.hbase.util.ReflectionUtils;
 import org.apache.hadoop.hbase.util.RetryCounter;
 import org.apache.hadoop.hbase.util.RetryCounterFactory;
 import org.apache.hadoop.hbase.util.TableDescriptorChecker;
@@ -446,6 +448,15 @@ public class HMaster extends HRegionServer implements MasterServices {
   public static final String WARMUP_BEFORE_MOVE = "hbase.master.warmup.before.move";
   private static final boolean DEFAULT_WARMUP_BEFORE_MOVE = true;
 
+  /**
+   * Use RSProcedureDispatcher instance to initiate master to rs remote procedure execution. Use
+   * this config to extend RSProcedureDispatcher (mainly for testing purpose).
+   */
+  public static final String HBASE_MASTER_RSPROC_DISPATCHER_CLASS =
+    "hbase.master.rsproc.dispatcher.class";
+  private static final String DEFAULT_HBASE_MASTER_RSPROC_DISPATCHER_CLASS =
+    RSProcedureDispatcher.class.getName();
+
   private TaskGroup startupTaskGroup;
 
   /**
@@ -1736,7 +1747,11 @@ protected void stopServiceThreads() {
   }
 
   private void createProcedureExecutor() throws IOException {
-    MasterProcedureEnv procEnv = new MasterProcedureEnv(this);
+    final String procedureDispatcherClassName =
+      conf.get(HBASE_MASTER_RSPROC_DISPATCHER_CLASS, DEFAULT_HBASE_MASTER_RSPROC_DISPATCHER_CLASS);
+    final RSProcedureDispatcher procedureDispatcher = ReflectionUtils.instantiateWithCustomCtor(
+      procedureDispatcherClassName, new Class[] { MasterServices.class }, new Object[] { this });
+    final MasterProcedureEnv procEnv = new MasterProcedureEnv(this, procedureDispatcher);
     procedureStore = new RegionProcedureStore(this, masterRegion,
       new MasterProcedureEnv.FsUtilsLeaseRecovery(this));
     procedureStore.registerListener(new ProcedureStoreListener() {

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

@@ -74,10 +74,6 @@ private boolean isRunning() {
   private final MasterProcedureScheduler procSched;
   private final MasterServices master;
 
-  public MasterProcedureEnv(final MasterServices master) {
-    this(master, new RSProcedureDispatcher(master));
-  }
-
   public MasterProcedureEnv(final MasterServices master,
     final RSProcedureDispatcher remoteDispatcher) {
     this.master = master;

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

@@ -27,14 +27,14 @@
 import org.apache.hadoop.hbase.DoNotRetryIOException;
 import org.apache.hadoop.hbase.ServerName;
 import org.apache.hadoop.hbase.client.RegionInfo;
+import org.apache.hadoop.hbase.exceptions.ConnectionClosedException;
 import org.apache.hadoop.hbase.ipc.RpcConnectionConstants;
 import org.apache.hadoop.hbase.ipc.ServerNotRunningYetException;
 import org.apache.hadoop.hbase.master.MasterServices;
 import org.apache.hadoop.hbase.master.ServerListener;
 import org.apache.hadoop.hbase.master.ServerManager;
 import org.apache.hadoop.hbase.procedure2.ProcedureExecutor;
 import org.apache.hadoop.hbase.procedure2.RemoteProcedureDispatcher;
-import org.apache.hadoop.hbase.regionserver.RegionServerAbortedException;
 import org.apache.hadoop.hbase.regionserver.RegionServerStoppedException;
 import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
 import org.apache.hadoop.ipc.RemoteException;
@@ -249,6 +249,22 @@ protected class ExecuteProceduresRemoteCall implements RemoteProcedureResolver,
       "hbase.regionserver.rpc.retry.interval";
     private static final int DEFAULT_RS_RPC_RETRY_INTERVAL = 100;
 
+    /**
+     * Config to determine the retry limit while executing remote regionserver procedure. This retry
+     * limit applies to only specific errors. These errors could potentially get the remote
+     * procedure stuck for several minutes unless the retry limit is applied.
+     */
+    private static final String RS_REMOTE_PROC_FAIL_FAST_LIMIT =
+      "hbase.master.rs.remote.proc.fail.fast.limit";
+    /**
+     * The default retry limit. Waiting for more than {@value} attempts is not going to help much
+     * for genuine connectivity errors. Therefore, consider fail-fast after {@value} retries. Value
+     * = {@value}
+     */
+    private static final int DEFAULT_RS_REMOTE_PROC_RETRY_LIMIT = 5;
+
+    private final int failFastRetryLimit;
+
     private ExecuteProceduresRequest.Builder request = null;
 
     public ExecuteProceduresRemoteCall(final ServerName serverName,
@@ -257,6 +273,8 @@ public ExecuteProceduresRemoteCall(final ServerName serverName,
       this.remoteProcedures = remoteProcedures;
       this.rsRpcRetryInterval = master.getConfiguration().getLong(RS_RPC_RETRY_INTERVAL_CONF_KEY,
         DEFAULT_RS_RPC_RETRY_INTERVAL);
+      this.failFastRetryLimit = master.getConfiguration().getInt(RS_REMOTE_PROC_FAIL_FAST_LIMIT,
+        DEFAULT_RS_REMOTE_PROC_RETRY_LIMIT);
     }
 
     private AdminService.BlockingInterface getRsAdmin() throws IOException {
@@ -305,13 +323,28 @@ private boolean scheduleForRetry(IOException e) {
       if (numberOfAttemptsSoFar == 0 && unableToConnectToServer(e)) {
         return false;
       }
+
+      // Check if the num of attempts have crossed the retry limit, and if the error type can
+      // fail-fast.
+      if (numberOfAttemptsSoFar >= failFastRetryLimit - 1 && isErrorTypeFailFast(e)) {
+        LOG
+          .warn("Number of retries {} exceeded limit {} for the given error type. Scheduling server"
+            + " crash for {}", numberOfAttemptsSoFar + 1, failFastRetryLimit, serverName, e);
+        // Expiring the server will schedule SCP and also reject the regionserver report from the
+        // regionserver if regionserver is somehow able to send the regionserver report to master.
+        // The master rejects the report by throwing YouAreDeadException, which would eventually
+        // result in the regionserver abort.
+        // This will also remove "serverName" from the ServerManager's onlineServers map.
+        master.getServerManager().expireServer(serverName);
+        return false;
+      }
       // Always retry for other exception types if the region server is not dead yet.
       if (!master.getServerManager().isServerOnline(serverName)) {
         LOG.warn("Request to {} failed due to {}, try={} and the server is not online, give up",
           serverName, e.toString(), numberOfAttemptsSoFar);
         return false;
       }
-      if (e instanceof RegionServerAbortedException || e instanceof RegionServerStoppedException) {
+      if (e instanceof RegionServerStoppedException) {
         // A better way is to return true here to let the upper layer quit, and then schedule a
         // background task to check whether the region server is dead. And if it is dead, call
         // remoteCallFailed to tell the upper layer. Keep retrying here does not lead to incorrect
@@ -329,7 +362,8 @@ private boolean scheduleForRetry(IOException e) {
       // retry^2 on each try
       // up to max of 10 seconds (don't want to back off too much in case of situation change).
       submitTask(this,
-        Math.min(rsRpcRetryInterval * (this.numberOfAttemptsSoFar * this.numberOfAttemptsSoFar),
+        Math.min(
+          rsRpcRetryInterval * ((long) this.numberOfAttemptsSoFar * this.numberOfAttemptsSoFar),
           10 * 1000),
         TimeUnit.MILLISECONDS);
       return true;
@@ -376,6 +410,39 @@ private boolean isSaslError(IOException e) {
       }
     }
 
+    /**
+     * Returns true if the error or its cause is of type ConnectionClosedException.
+     * @param e IOException thrown by the underlying rpc framework.
+     * @return True if the error or its cause is of type ConnectionClosedException.
+     */
+    private boolean isConnectionClosedError(IOException e) {
+      if (e instanceof ConnectionClosedException) {
+        return true;
+      }
+      Throwable cause = e;
+      while (true) {
+        if (cause instanceof IOException) {
+          IOException unwrappedCause = unwrapException((IOException) cause);
+          if (unwrappedCause instanceof ConnectionClosedException) {
+            return true;
+          }
+        }
+        cause = cause.getCause();
+        if (cause == null) {
+          return false;
+        }
+      }
+    }
+
+    /**
+     * Returns true if the error type can allow fail-fast.
+     * @param e IOException thrown by the underlying rpc framework.
+     * @return True if the error type can allow fail-fast.
+     */
+    private boolean isErrorTypeFailFast(IOException e) {
+      return e instanceof CallQueueTooBigException || isSaslError(e) || isConnectionClosedError(e);
+    }
+
     private long getMaxWaitTime() {
       if (this.maxWaitTime < 0) {
         // This is the max attempts, not retries, so it should be at least 1.

hbase-server/src/test/java/org/apache/hadoop/hbase/master/assignment/TestAssignmentManagerBase.java

@@ -146,6 +146,7 @@ protected void setupConfiguration(Configuration conf) throws Exception {
     // make retry for TRSP more frequent
     conf.setLong(ProcedureUtil.PROCEDURE_RETRY_SLEEP_INTERVAL_MS, 10);
     conf.setLong(ProcedureUtil.PROCEDURE_RETRY_MAX_SLEEP_TIME_MS, 100);
+    conf.setInt("hbase.master.rs.remote.proc.fail.fast.limit", Integer.MAX_VALUE);
   }
 
   @Before

hbase-server/src/test/java/org/apache/hadoop/hbase/util/RSProcDispatcher.java

@@ -0,0 +1,114 @@
+/*
+ * 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
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.hbase.util;
+
+import java.io.IOException;
+import java.util.Set;
+import java.util.concurrent.atomic.AtomicInteger;
+import org.apache.hadoop.hbase.ServerName;
+import org.apache.hadoop.hbase.exceptions.ConnectionClosedException;
+import org.apache.hadoop.hbase.master.MasterServices;
+import org.apache.hadoop.hbase.master.procedure.RSProcedureDispatcher;
+import org.apache.hadoop.hbase.regionserver.RegionServerStoppedException;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.hbase.thirdparty.com.google.protobuf.ServiceException;
+
+import org.apache.hadoop.hbase.shaded.protobuf.generated.AdminProtos;
+
+/**
+ * Test implementation of RSProcedureDispatcher that throws desired errors for testing purpose.
+ */
+public class RSProcDispatcher extends RSProcedureDispatcher {
+
+  private static final Logger LOG = LoggerFactory.getLogger(RSProcDispatcher.class);
+
+  private static final AtomicInteger i = new AtomicInteger();
+
+  public RSProcDispatcher(MasterServices master) {
+    super(master);
+  }
+
+  @Override
+  protected void remoteDispatch(final ServerName serverName,
+    final Set<RemoteProcedure> remoteProcedures) {
+    if (!master.getServerManager().isServerOnline(serverName)) {
+      // fail fast
+      submitTask(new DeadRSRemoteCall(serverName, remoteProcedures));
+    } else {
+      submitTask(new TestExecuteProceduresRemoteCall(serverName, remoteProcedures));
+    }
+  }
+
+  class TestExecuteProceduresRemoteCall extends ExecuteProceduresRemoteCall {
+
+    public TestExecuteProceduresRemoteCall(ServerName serverName,
+      Set<RemoteProcedure> remoteProcedures) {
+      super(serverName, remoteProcedures);
+    }
+
+    @Override
+    public AdminProtos.ExecuteProceduresResponse sendRequest(final ServerName serverName,
+      final AdminProtos.ExecuteProceduresRequest request) throws IOException {
+      int j = i.addAndGet(1);
+      LOG.info("sendRequest() req: {} , j: {}", request, j);
+      if (j == 12 || j == 22) {
+        // Execute the remote close and open region requests in the last (5th) retry before
+        // throwing ConnectionClosedException. This is to ensure even if the region open/close
+        // is successfully completed by regionserver, master still schedules SCP because
+        // sendRequest() throws error which has retry-limit exhausted.
+        try {
+          getRsAdmin().executeProcedures(null, request);
+        } catch (ServiceException e) {
+          throw new RuntimeException(e);
+        }
+      }
+      // For one of the close region requests and one of the open region requests,
+      // throw ConnectionClosedException until retry limit is exhausted and master
+      // schedules recoveries for the server.
+      // We will have ABNORMALLY_CLOSED regions, and they are expected to recover on their own.
+      if (j >= 10 && j <= 15 || j >= 18 && j <= 23) {
+        throw new ConnectionClosedException("test connection closed error...");
+      }
+      try {
+        return getRsAdmin().executeProcedures(null, request);
+      } catch (ServiceException e) {
+        throw new RuntimeException(e);
+      }
+    }
+
+    private AdminProtos.AdminService.BlockingInterface getRsAdmin() throws IOException {
+      return master.getServerManager().getRsAdmin(getServerName());
+    }
+  }
+
+  private class DeadRSRemoteCall extends ExecuteProceduresRemoteCall {
+
+    public DeadRSRemoteCall(ServerName serverName, Set<RemoteProcedure> remoteProcedures) {
+      super(serverName, remoteProcedures);
+    }
+
+    @Override
+    public void run() {
+      remoteCallFailed(master.getMasterProcedureExecutor().getEnvironment(),
+        new RegionServerStoppedException("Server " + getServerName() + " is not online"));
+    }
+  }
+
+}