ratis-grpc/src/main/java/org/apache/ratis/grpc/server/GRpcLogAppender.java - ratis-hadoop-projects - Git at Google

 /**
  * 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.ratis.grpc.server;

 import org.apache.ratis.grpc.GrpcConfigKeys;
 import org.apache.ratis.grpc.RaftGRpcService;
 import org.apache.ratis.grpc.RaftGrpcUtil;
 import org.apache.ratis.server.RaftServerConfigKeys;
 import org.apache.ratis.server.impl.FollowerInfo;
 import org.apache.ratis.server.impl.LeaderState;
 import org.apache.ratis.server.impl.LogAppender;
 import org.apache.ratis.server.impl.RaftServerImpl;
 import org.apache.ratis.shaded.io.grpc.stub.StreamObserver;
 import org.apache.ratis.shaded.proto.RaftProtos.AppendEntriesReplyProto;
 import org.apache.ratis.shaded.proto.RaftProtos.AppendEntriesRequestProto;
 import org.apache.ratis.shaded.proto.RaftProtos.InstallSnapshotReplyProto;
 import org.apache.ratis.shaded.proto.RaftProtos.InstallSnapshotRequestProto;
 import org.apache.ratis.statemachine.SnapshotInfo;
 import org.apache.ratis.util.*;

 import java.io.IOException;
 import java.util.*;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.atomic.AtomicBoolean;

 /**
  * A new log appender implementation using grpc bi-directional stream API.
  */
 public class GRpcLogAppender extends LogAppender {
   private final RaftGRpcService rpcService;
   private final Map<Long, AppendEntriesRequestProto> pendingRequests;
   private final int maxPendingRequestsNum;
   private long callId = 0;
   private volatile boolean firstResponseReceived = false;

   private final TimeDuration requestTimeoutDuration;
   private final TimeoutScheduler scheduler = TimeoutScheduler.newInstance(1);

   private volatile StreamObserver<AppendEntriesRequestProto> appendLogRequestObserver;

   public GRpcLogAppender(RaftServerImpl server, LeaderState leaderState,
                          FollowerInfo f) {
     super(server, leaderState, f);

     this.rpcService = (RaftGRpcService) server.getServerRpc();

     maxPendingRequestsNum = GrpcConfigKeys.Server.leaderOutstandingAppendsMax(
         server.getProxy().getProperties());
     requestTimeoutDuration = RaftServerConfigKeys.Rpc.requestTimeout(server.getProxy().getProperties());
     pendingRequests = new ConcurrentHashMap<>();
   }

   private RaftServerProtocolClient getClient() throws IOException {
     return rpcService.getProxies().getProxy(follower.getPeer().getId());
   }

   private synchronized void resetClient(AppendEntriesRequestProto request) {
     rpcService.getProxies().resetProxy(follower.getPeer().getId());
     appendLogRequestObserver = null;
     firstResponseReceived = false;

     // clear the pending requests queue and reset the next index of follower
     final long nextIndex = request != null && request.hasPreviousLog()?
         request.getPreviousLog().getIndex() + 1: raftLog.getStartIndex();
     clearPendingRequests(nextIndex);
   }

   @Override
   public void run() {
     for(; isAppenderRunning(); mayWait()) {
       if (shouldSendRequest()) {
         SnapshotInfo snapshot = shouldInstallSnapshot();
         if (snapshot != null) {
           installSnapshot(snapshot);
         } else if (!shouldWait()) {
           // keep appending log entries or sending heartbeats
           try {
             appendLog();
           } catch (IOException e) {
             LOG.error(this + " hit IOException while loading raft log", e);
             stopSender();
             return;
           }
         }
       }
     }

     Optional.ofNullable(appendLogRequestObserver).ifPresent(StreamObserver::onCompleted);
   }

   private long getWaitTimeMs() {
     if (!shouldSendRequest()) {
       return getHeartbeatRemainingTime(); // No requests, wait until heartbeat
     } else if (shouldWait()) {
       return halfMinTimeoutMs; // Should wait for a short time
     }
     return 0L;
   }

   private void mayWait() {
     // use lastSend time instead of lastResponse time
     final long waitTimeMs = getWaitTimeMs();
     if (waitTimeMs <= 0L) {
       return;
     }

     synchronized(this) {
       try {
         LOG.trace("{}: wait {}ms", this, waitTimeMs);
         wait(waitTimeMs);
       } catch(InterruptedException ie) {
         LOG.warn(this + ": Wait interrupted by " + ie);
       }
     }
   }

   @Override
   protected boolean shouldSendRequest() {
     return appendLogRequestObserver == null || super.shouldSendRequest();
   }

   /** @return true iff not received first response or queue is full. */
   private boolean shouldWait() {
     final int size = pendingRequests.size();
     if (size == 0) {
       return false;
     }
     return !firstResponseReceived || size >= maxPendingRequestsNum;
   }

   private void appendLog() throws IOException {
     final AppendEntriesRequestProto pending;
     final StreamObserver<AppendEntriesRequestProto> s;
     synchronized (this) {
       // prepare and enqueue the append request. note changes on follower's
       // nextIndex and ops on pendingRequests should always be associated
       // together and protected by the lock
       pending = createRequest(callId++);
       if (pending == null) {
         return;
       }
       pendingRequests.put(pending.getServerRequest().getCallId(), pending);
       updateNextIndex(pending);
       if (appendLogRequestObserver == null) {
         appendLogRequestObserver = getClient().appendEntries(new AppendLogResponseHandler());
       }
       s = appendLogRequestObserver;
     }

     if (isAppenderRunning()) {
       sendRequest(pending, s);
     }
   }

   private void sendRequest(AppendEntriesRequestProto request,
       StreamObserver<AppendEntriesRequestProto> s) {
     CodeInjectionForTesting.execute(RaftGRpcService.GRPC_SEND_SERVER_REQUEST,
         server.getId(), null, request);

     s.onNext(request);
     scheduler.onTimeout(requestTimeoutDuration, () -> timeoutAppendRequest(request), LOG,
         () -> "Timeout check failed for append entry request: " + request);
     follower.updateLastRpcSendTime();
   }

   private void timeoutAppendRequest(AppendEntriesRequestProto request) {
     AppendEntriesRequestProto pendingRequest = pendingRequests.remove(request.getServerRequest().getCallId());
     if (pendingRequest != null) {
       final String err = this + ": appendEntries Timeout, request=" + ProtoUtils.toString(pendingRequest.getServerRequest());
       LOG.warn(err);
     }
   }

   private void updateNextIndex(AppendEntriesRequestProto request) {
     final int count = request.getEntriesCount();
     if (count > 0) {
       follower.updateNextIndex(request.getEntries(count - 1).getIndex() + 1);
     }
   }

   /**
    * StreamObserver for handling responses from the follower
    */
   private class AppendLogResponseHandler
       implements StreamObserver<AppendEntriesReplyProto> {
     /**
      * After receiving a appendEntries reply, do the following:
      * 1. If the reply is success, update the follower's match index and submit
      *    an event to leaderState
      * 2. If the reply is NOT_LEADER, step down
      * 3. If the reply is INCONSISTENCY, decrease the follower's next index
      *    based on the response
      */
     @Override
     public void onNext(AppendEntriesReplyProto reply) {
       LOG.debug("{} received {} response from {}", server.getId(),
           (!firstResponseReceived ? "the first" : "a"),
           follower.getPeer());

       // update the last rpc time
       follower.updateLastRpcResponseTime();

       if (!firstResponseReceived) {
         firstResponseReceived = true;
       }
       switch (reply.getResult()) {
         case SUCCESS:
           onSuccess(reply);
           break;
         case NOT_LEADER:
           onNotLeader(reply);
           break;
         case INCONSISTENCY:
           onInconsistency(reply);
           break;
         default:
           break;
       }
       notifyAppend();
     }

     /**
      * for now we simply retry the first pending request
      */
     @Override
     public void onError(Throwable t) {
       if (!isAppenderRunning()) {
         LOG.info("{} is stopped", GRpcLogAppender.this);
         return;
       }
       RaftGrpcUtil.warn(LOG, () -> server.getId() + ": Failed appendEntries to " + follower.getPeer(), t);

       long callId = RaftGrpcUtil.getCallId(t);
       resetClient(pendingRequests.get(callId));
     }

     @Override
     public void onCompleted() {
       LOG.info("{} stops appending log entries to follower {}", server.getId(),
           follower);
     }
   }

   private void clearPendingRequests(long newNextIndex) {
     pendingRequests.clear();
     follower.decreaseNextIndex(newNextIndex);
   }

   protected synchronized void onSuccess(AppendEntriesReplyProto reply) {
     AppendEntriesRequestProto request = pendingRequests.remove(reply.getServerReply().getCallId());
     if (request == null) {
       // If reply comes after timeout, the reply is ignored.
       LOG.warn("Ignoring reply: " + reply);
       return;
     }
     updateCommitIndex(request.getLeaderCommit());

     final long replyNextIndex = reply.getNextIndex();
     Objects.requireNonNull(request,
         () -> "Got reply with next index " + replyNextIndex
             + " but the pending queue is empty");

     final long lastIndex = replyNextIndex - 1;
     final boolean updateMatchIndex;

     if (request.getEntriesCount() == 0) {
       Preconditions.assertTrue(!request.hasPreviousLog() ||
               lastIndex == request.getPreviousLog().getIndex(),
           "reply's next index is %s, request's previous is %s",
           replyNextIndex, request.getPreviousLog());
       updateMatchIndex = request.hasPreviousLog() && follower.getMatchIndex() < lastIndex;
     } else {
       // check if the reply and the pending request is consistent
       final long lastEntryIndex = request
           .getEntries(request.getEntriesCount() - 1).getIndex();
       Preconditions.assertTrue(lastIndex == lastEntryIndex,
           "reply's next index is %s, request's last entry index is %s",
           replyNextIndex, lastEntryIndex);
       updateMatchIndex = true;
     }
     if (updateMatchIndex) {
       follower.updateMatchIndex(lastIndex);
       submitEventOnSuccessAppend();
     }
   }

   private void onNotLeader(AppendEntriesReplyProto reply) {
     checkResponseTerm(reply.getTerm());
     // the running loop will end and the connection will onComplete
   }

   private synchronized void onInconsistency(AppendEntriesReplyProto reply) {
     AppendEntriesRequestProto request = pendingRequests.remove(reply.getServerReply().getCallId());
     if (request == null) {
       // If reply comes after timeout, the reply is ignored.
       LOG.warn("Ignoring reply: " + reply);
       return;
     }
     Preconditions.assertTrue(request.hasPreviousLog());
     if (request.getPreviousLog().getIndex() >= reply.getNextIndex()) {
       clearPendingRequests(reply.getNextIndex());
     }
   }

   private class InstallSnapshotResponseHandler
       implements StreamObserver<InstallSnapshotReplyProto> {
     private final Queue<Integer> pending;
     private final AtomicBoolean done = new AtomicBoolean(false);

     InstallSnapshotResponseHandler() {
       pending = new LinkedList<>();
     }

     synchronized void addPending(InstallSnapshotRequestProto request) {
       pending.offer(request.getRequestIndex());
     }

     synchronized void removePending(InstallSnapshotReplyProto reply) {
       int index = pending.poll();
       Preconditions.assertTrue(index == reply.getRequestIndex());
     }

     boolean isDone() {
       return done.get();
     }

     void close() {
       done.set(true);
       GRpcLogAppender.this.notifyAppend();
     }

     synchronized boolean hasAllResponse() {
       return pending.isEmpty();
     }

     @Override
     public void onNext(InstallSnapshotReplyProto reply) {
       LOG.debug("{} received {} response from {}", server.getId(),
           (!firstResponseReceived ? "the first" : "a"),
           follower.getPeer());

       // update the last rpc time
       follower.updateLastRpcResponseTime();

       if (!firstResponseReceived) {
         firstResponseReceived = true;
       }

       switch (reply.getResult()) {
         case SUCCESS:
           removePending(reply);
           break;
         case NOT_LEADER:
           checkResponseTerm(reply.getTerm());
           break;
         case UNRECOGNIZED:
           break;
       }
     }

     @Override
     public void onError(Throwable t) {
       if (!isAppenderRunning()) {
         LOG.info("{} is stopped", GRpcLogAppender.this);
         return;
       }
       LOG.info("{} got error when installing snapshot to {}, exception: {}",
           server.getId(), follower.getPeer(), t);
       resetClient(null);
       close();
     }

     @Override
     public void onCompleted() {
       LOG.info("{} stops sending snapshots to follower {}", server.getId(),
           follower);
       close();
     }
   }

   private void installSnapshot(SnapshotInfo snapshot) {
     LOG.info("{}: follower {}'s next index is {}," +
             " log's start index is {}, need to install snapshot",
         server.getId(), follower.getPeer(), follower.getNextIndex(),
         raftLog.getStartIndex());

     final InstallSnapshotResponseHandler responseHandler = new InstallSnapshotResponseHandler();
     StreamObserver<InstallSnapshotRequestProto> snapshotRequestObserver = null;
     final String requestId = UUID.randomUUID().toString();
     try {
       snapshotRequestObserver = getClient().installSnapshot(responseHandler);
       for (InstallSnapshotRequestProto request :
           new SnapshotRequestIter(snapshot, requestId)) {
         if (isAppenderRunning()) {
           snapshotRequestObserver.onNext(request);
           follower.updateLastRpcSendTime();
           responseHandler.addPending(request);
         } else {
           break;
         }
       }
       snapshotRequestObserver.onCompleted();
     } catch (Exception e) {
       LOG.warn("{} failed to install snapshot {}. Exception: {}", this,
           snapshot.getFiles(), e);
       if (snapshotRequestObserver != null) {
         snapshotRequestObserver.onError(e);
       }
       return;
     }

     synchronized (this) {
       while (isAppenderRunning() && !responseHandler.isDone()) {
         try {
           wait();
         } catch (InterruptedException ignored) {
         }
       }
     }

     if (responseHandler.hasAllResponse()) {
       follower.updateMatchIndex(snapshot.getTermIndex().getIndex());
       follower.updateNextIndex(snapshot.getTermIndex().getIndex() + 1);
       LOG.info("{}: install snapshot-{} successfully on follower {}",
           server.getId(), snapshot.getTermIndex().getIndex(), follower.getPeer());
     }
   }
 }
	/**
	* 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.ratis.grpc.server;

	import org.apache.ratis.grpc.GrpcConfigKeys;
	import org.apache.ratis.grpc.RaftGRpcService;
	import org.apache.ratis.grpc.RaftGrpcUtil;
	import org.apache.ratis.server.RaftServerConfigKeys;
	import org.apache.ratis.server.impl.FollowerInfo;
	import org.apache.ratis.server.impl.LeaderState;
	import org.apache.ratis.server.impl.LogAppender;
	import org.apache.ratis.server.impl.RaftServerImpl;
	import org.apache.ratis.shaded.io.grpc.stub.StreamObserver;
	import org.apache.ratis.shaded.proto.RaftProtos.AppendEntriesReplyProto;
	import org.apache.ratis.shaded.proto.RaftProtos.AppendEntriesRequestProto;
	import org.apache.ratis.shaded.proto.RaftProtos.InstallSnapshotReplyProto;
	import org.apache.ratis.shaded.proto.RaftProtos.InstallSnapshotRequestProto;
	import org.apache.ratis.statemachine.SnapshotInfo;
	import org.apache.ratis.util.*;

	import java.io.IOException;
	import java.util.*;
	import java.util.concurrent.ConcurrentHashMap;
	import java.util.concurrent.atomic.AtomicBoolean;

	/**
	* A new log appender implementation using grpc bi-directional stream API.
	*/
	public class GRpcLogAppender extends LogAppender {
	private final RaftGRpcService rpcService;
	private final Map<Long, AppendEntriesRequestProto> pendingRequests;
	private final int maxPendingRequestsNum;
	private long callId = 0;
	private volatile boolean firstResponseReceived = false;

	private final TimeDuration requestTimeoutDuration;
	private final TimeoutScheduler scheduler = TimeoutScheduler.newInstance(1);

	private volatile StreamObserver<AppendEntriesRequestProto> appendLogRequestObserver;

	public GRpcLogAppender(RaftServerImpl server, LeaderState leaderState,
	FollowerInfo f) {
	super(server, leaderState, f);

	this.rpcService = (RaftGRpcService) server.getServerRpc();

	maxPendingRequestsNum = GrpcConfigKeys.Server.leaderOutstandingAppendsMax(
	server.getProxy().getProperties());
	requestTimeoutDuration = RaftServerConfigKeys.Rpc.requestTimeout(server.getProxy().getProperties());
	pendingRequests = new ConcurrentHashMap<>();
	}

	private RaftServerProtocolClient getClient() throws IOException {
	return rpcService.getProxies().getProxy(follower.getPeer().getId());
	}

	private synchronized void resetClient(AppendEntriesRequestProto request) {
	rpcService.getProxies().resetProxy(follower.getPeer().getId());
	appendLogRequestObserver = null;
	firstResponseReceived = false;

	// clear the pending requests queue and reset the next index of follower
	final long nextIndex = request != null && request.hasPreviousLog()?
	request.getPreviousLog().getIndex() + 1: raftLog.getStartIndex();
	clearPendingRequests(nextIndex);
	}

	@Override
	public void run() {
	for(; isAppenderRunning(); mayWait()) {
	if (shouldSendRequest()) {
	SnapshotInfo snapshot = shouldInstallSnapshot();
	if (snapshot != null) {
	installSnapshot(snapshot);
	} else if (!shouldWait()) {
	// keep appending log entries or sending heartbeats
	try {
	appendLog();
	} catch (IOException e) {
	LOG.error(this + " hit IOException while loading raft log", e);
	stopSender();
	return;
	}
	}
	}
	}

	Optional.ofNullable(appendLogRequestObserver).ifPresent(StreamObserver::onCompleted);
	}

	private long getWaitTimeMs() {
	if (!shouldSendRequest()) {
	return getHeartbeatRemainingTime(); // No requests, wait until heartbeat
	} else if (shouldWait()) {
	return halfMinTimeoutMs; // Should wait for a short time
	}
	return 0L;
	}

	private void mayWait() {
	// use lastSend time instead of lastResponse time
	final long waitTimeMs = getWaitTimeMs();
	if (waitTimeMs <= 0L) {
	return;
	}

	synchronized(this) {
	try {
	LOG.trace("{}: wait {}ms", this, waitTimeMs);
	wait(waitTimeMs);
	} catch(InterruptedException ie) {
	LOG.warn(this + ": Wait interrupted by " + ie);
	}
	}
	}

	@Override
	protected boolean shouldSendRequest() {
	return appendLogRequestObserver == null \|\| super.shouldSendRequest();
	}

	/** @return true iff not received first response or queue is full. */
	private boolean shouldWait() {
	final int size = pendingRequests.size();
	if (size == 0) {
	return false;
	}
	return !firstResponseReceived \|\| size >= maxPendingRequestsNum;
	}

	private void appendLog() throws IOException {
	final AppendEntriesRequestProto pending;
	final StreamObserver<AppendEntriesRequestProto> s;
	synchronized (this) {
	// prepare and enqueue the append request. note changes on follower's
	// nextIndex and ops on pendingRequests should always be associated
	// together and protected by the lock
	pending = createRequest(callId++);
	if (pending == null) {
	return;
	}
	pendingRequests.put(pending.getServerRequest().getCallId(), pending);
	updateNextIndex(pending);
	if (appendLogRequestObserver == null) {
	appendLogRequestObserver = getClient().appendEntries(new AppendLogResponseHandler());
	}
	s = appendLogRequestObserver;
	}

	if (isAppenderRunning()) {
	sendRequest(pending, s);
	}
	}

	private void sendRequest(AppendEntriesRequestProto request,
	StreamObserver<AppendEntriesRequestProto> s) {
	CodeInjectionForTesting.execute(RaftGRpcService.GRPC_SEND_SERVER_REQUEST,
	server.getId(), null, request);

	s.onNext(request);
	scheduler.onTimeout(requestTimeoutDuration, () -> timeoutAppendRequest(request), LOG,
	() -> "Timeout check failed for append entry request: " + request);
	follower.updateLastRpcSendTime();
	}

	private void timeoutAppendRequest(AppendEntriesRequestProto request) {
	AppendEntriesRequestProto pendingRequest = pendingRequests.remove(request.getServerRequest().getCallId());
	if (pendingRequest != null) {
	final String err = this + ": appendEntries Timeout, request=" + ProtoUtils.toString(pendingRequest.getServerRequest());
	LOG.warn(err);
	}
	}

	private void updateNextIndex(AppendEntriesRequestProto request) {
	final int count = request.getEntriesCount();
	if (count > 0) {
	follower.updateNextIndex(request.getEntries(count - 1).getIndex() + 1);
	}
	}

	/**
	* StreamObserver for handling responses from the follower
	*/
	private class AppendLogResponseHandler
	implements StreamObserver<AppendEntriesReplyProto> {
	/**
	* After receiving a appendEntries reply, do the following:
	* 1. If the reply is success, update the follower's match index and submit
	* an event to leaderState
	* 2. If the reply is NOT_LEADER, step down
	* 3. If the reply is INCONSISTENCY, decrease the follower's next index
	* based on the response
	*/
	@Override
	public void onNext(AppendEntriesReplyProto reply) {
	LOG.debug("{} received {} response from {}", server.getId(),
	(!firstResponseReceived ? "the first" : "a"),
	follower.getPeer());

	// update the last rpc time
	follower.updateLastRpcResponseTime();

	if (!firstResponseReceived) {
	firstResponseReceived = true;
	}
	switch (reply.getResult()) {
	case SUCCESS:
	onSuccess(reply);
	break;
	case NOT_LEADER:
	onNotLeader(reply);
	break;
	case INCONSISTENCY:
	onInconsistency(reply);
	break;
	default:
	break;
	}
	notifyAppend();
	}

	/**
	* for now we simply retry the first pending request
	*/
	@Override
	public void onError(Throwable t) {
	if (!isAppenderRunning()) {
	LOG.info("{} is stopped", GRpcLogAppender.this);
	return;
	}
	RaftGrpcUtil.warn(LOG, () -> server.getId() + ": Failed appendEntries to " + follower.getPeer(), t);

	long callId = RaftGrpcUtil.getCallId(t);
	resetClient(pendingRequests.get(callId));
	}

	@Override
	public void onCompleted() {
	LOG.info("{} stops appending log entries to follower {}", server.getId(),
	follower);
	}
	}

	private void clearPendingRequests(long newNextIndex) {
	pendingRequests.clear();
	follower.decreaseNextIndex(newNextIndex);
	}

	protected synchronized void onSuccess(AppendEntriesReplyProto reply) {
	AppendEntriesRequestProto request = pendingRequests.remove(reply.getServerReply().getCallId());
	if (request == null) {
	// If reply comes after timeout, the reply is ignored.
	LOG.warn("Ignoring reply: " + reply);
	return;
	}
	updateCommitIndex(request.getLeaderCommit());

	final long replyNextIndex = reply.getNextIndex();
	Objects.requireNonNull(request,
	() -> "Got reply with next index " + replyNextIndex
	+ " but the pending queue is empty");

	final long lastIndex = replyNextIndex - 1;
	final boolean updateMatchIndex;

	if (request.getEntriesCount() == 0) {
	Preconditions.assertTrue(!request.hasPreviousLog() \|\|
	lastIndex == request.getPreviousLog().getIndex(),
	"reply's next index is %s, request's previous is %s",
	replyNextIndex, request.getPreviousLog());
	updateMatchIndex = request.hasPreviousLog() && follower.getMatchIndex() < lastIndex;
	} else {
	// check if the reply and the pending request is consistent
	final long lastEntryIndex = request
	.getEntries(request.getEntriesCount() - 1).getIndex();
	Preconditions.assertTrue(lastIndex == lastEntryIndex,
	"reply's next index is %s, request's last entry index is %s",
	replyNextIndex, lastEntryIndex);
	updateMatchIndex = true;
	}
	if (updateMatchIndex) {
	follower.updateMatchIndex(lastIndex);
	submitEventOnSuccessAppend();
	}
	}

	private void onNotLeader(AppendEntriesReplyProto reply) {
	checkResponseTerm(reply.getTerm());
	// the running loop will end and the connection will onComplete
	}

	private synchronized void onInconsistency(AppendEntriesReplyProto reply) {
	AppendEntriesRequestProto request = pendingRequests.remove(reply.getServerReply().getCallId());
	if (request == null) {
	// If reply comes after timeout, the reply is ignored.
	LOG.warn("Ignoring reply: " + reply);
	return;
	}
	Preconditions.assertTrue(request.hasPreviousLog());
	if (request.getPreviousLog().getIndex() >= reply.getNextIndex()) {
	clearPendingRequests(reply.getNextIndex());
	}
	}

	private class InstallSnapshotResponseHandler
	implements StreamObserver<InstallSnapshotReplyProto> {
	private final Queue<Integer> pending;
	private final AtomicBoolean done = new AtomicBoolean(false);

	InstallSnapshotResponseHandler() {
	pending = new LinkedList<>();
	}

	synchronized void addPending(InstallSnapshotRequestProto request) {
	pending.offer(request.getRequestIndex());
	}

	synchronized void removePending(InstallSnapshotReplyProto reply) {
	int index = pending.poll();
	Preconditions.assertTrue(index == reply.getRequestIndex());
	}

	boolean isDone() {
	return done.get();
	}

	void close() {
	done.set(true);
	GRpcLogAppender.this.notifyAppend();
	}

	synchronized boolean hasAllResponse() {
	return pending.isEmpty();
	}

	@Override
	public void onNext(InstallSnapshotReplyProto reply) {
	LOG.debug("{} received {} response from {}", server.getId(),
	(!firstResponseReceived ? "the first" : "a"),
	follower.getPeer());

	// update the last rpc time
	follower.updateLastRpcResponseTime();

	if (!firstResponseReceived) {
	firstResponseReceived = true;
	}

	switch (reply.getResult()) {
	case SUCCESS:
	removePending(reply);
	break;
	case NOT_LEADER:
	checkResponseTerm(reply.getTerm());
	break;
	case UNRECOGNIZED:
	break;
	}
	}

	@Override
	public void onError(Throwable t) {
	if (!isAppenderRunning()) {
	LOG.info("{} is stopped", GRpcLogAppender.this);
	return;
	}
	LOG.info("{} got error when installing snapshot to {}, exception: {}",
	server.getId(), follower.getPeer(), t);
	resetClient(null);
	close();
	}

	@Override
	public void onCompleted() {
	LOG.info("{} stops sending snapshots to follower {}", server.getId(),
	follower);
	close();
	}
	}

	private void installSnapshot(SnapshotInfo snapshot) {
	LOG.info("{}: follower {}'s next index is {}," +
	" log's start index is {}, need to install snapshot",
	server.getId(), follower.getPeer(), follower.getNextIndex(),
	raftLog.getStartIndex());

	final InstallSnapshotResponseHandler responseHandler = new InstallSnapshotResponseHandler();
	StreamObserver<InstallSnapshotRequestProto> snapshotRequestObserver = null;
	final String requestId = UUID.randomUUID().toString();
	try {
	snapshotRequestObserver = getClient().installSnapshot(responseHandler);
	for (InstallSnapshotRequestProto request :
	new SnapshotRequestIter(snapshot, requestId)) {
	if (isAppenderRunning()) {
	snapshotRequestObserver.onNext(request);
	follower.updateLastRpcSendTime();
	responseHandler.addPending(request);
	} else {
	break;
	}
	}
	snapshotRequestObserver.onCompleted();
	} catch (Exception e) {
	LOG.warn("{} failed to install snapshot {}. Exception: {}", this,
	snapshot.getFiles(), e);
	if (snapshotRequestObserver != null) {
	snapshotRequestObserver.onError(e);
	}
	return;
	}

	synchronized (this) {
	while (isAppenderRunning() && !responseHandler.isDone()) {
	try {
	wait();
	} catch (InterruptedException ignored) {
	}
	}
	}

	if (responseHandler.hasAllResponse()) {
	follower.updateMatchIndex(snapshot.getTermIndex().getIndex());
	follower.updateNextIndex(snapshot.getTermIndex().getIndex() + 1);
	LOG.info("{}: install snapshot-{} successfully on follower {}",
	server.getId(), snapshot.getTermIndex().getIndex(), follower.getPeer());
	}
	}
	}