blob: 4ae62c1d0b64a70dcd06535e07b3c6960e7162f6 [file] [log] [blame]
/*
* 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.cassandra.net;
import java.io.DataInputStream;
import java.io.DataOutput;
import java.io.IOException;
import java.net.InetAddress;
import java.net.Socket;
import java.net.SocketException;
import java.nio.ByteBuffer;
import java.nio.channels.Channels;
import java.nio.channels.WritableByteChannel;
import java.util.*;
import java.util.concurrent.BlockingQueue;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.LinkedBlockingQueue;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
import java.util.zip.Checksum;
import javax.net.ssl.SSLHandshakeException;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import io.netty.util.concurrent.FastThreadLocalThread;
import net.jpountz.lz4.LZ4BlockOutputStream;
import net.jpountz.lz4.LZ4Compressor;
import net.jpountz.lz4.LZ4Factory;
import net.jpountz.xxhash.XXHashFactory;
import org.apache.cassandra.concurrent.NamedThreadFactory;
import org.apache.cassandra.io.util.DataOutputStreamPlus;
import org.apache.cassandra.io.util.BufferedDataOutputStreamPlus;
import org.apache.cassandra.io.util.WrappedDataOutputStreamPlus;
import org.apache.cassandra.tracing.TraceState;
import org.apache.cassandra.tracing.Tracing;
import org.apache.cassandra.utils.CoalescingStrategies;
import org.apache.cassandra.utils.CoalescingStrategies.Coalescable;
import org.apache.cassandra.utils.CoalescingStrategies.CoalescingStrategy;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.JVMStabilityInspector;
import org.apache.cassandra.utils.NanoTimeToCurrentTimeMillis;
import org.apache.cassandra.utils.UUIDGen;
import org.xerial.snappy.SnappyOutputStream;
import org.apache.cassandra.config.Config;
import org.apache.cassandra.config.DatabaseDescriptor;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.util.concurrent.Uninterruptibles;
public class OutboundTcpConnection extends FastThreadLocalThread
{
private static final Logger logger = LoggerFactory.getLogger(OutboundTcpConnection.class);
private static final String PREFIX = Config.PROPERTY_PREFIX;
/*
* Enabled/disable TCP_NODELAY for intradc connections. Defaults to enabled.
*/
private static final String INTRADC_TCP_NODELAY_PROPERTY = PREFIX + "otc_intradc_tcp_nodelay";
private static final boolean INTRADC_TCP_NODELAY = Boolean.parseBoolean(System.getProperty(INTRADC_TCP_NODELAY_PROPERTY, "true"));
/*
* Size of buffer in output stream
*/
private static final String BUFFER_SIZE_PROPERTY = PREFIX + "otc_buffer_size";
private static final int BUFFER_SIZE = Integer.getInteger(BUFFER_SIZE_PROPERTY, 1024 * 64);
public static final int MAX_COALESCED_MESSAGES = 128;
private static CoalescingStrategy newCoalescingStrategy(String displayName)
{
return CoalescingStrategies.newCoalescingStrategy(DatabaseDescriptor.getOtcCoalescingStrategy(),
DatabaseDescriptor.getOtcCoalescingWindow(),
logger,
displayName);
}
static
{
String strategy = DatabaseDescriptor.getOtcCoalescingStrategy();
switch (strategy)
{
case "TIMEHORIZON":
break;
case "MOVINGAVERAGE":
case "FIXED":
case "DISABLED":
logger.info("OutboundTcpConnection using coalescing strategy {}", strategy);
break;
default:
//Check that it can be loaded
newCoalescingStrategy("dummy");
}
int coalescingWindow = DatabaseDescriptor.getOtcCoalescingWindow();
if (coalescingWindow != Config.otc_coalescing_window_us_default)
logger.info("OutboundTcpConnection coalescing window set to {}μs", coalescingWindow);
if (coalescingWindow < 0)
throw new ExceptionInInitializerError(
"Value provided for coalescing window must be greater than 0: " + coalescingWindow);
int otc_backlog_expiration_interval_in_ms = DatabaseDescriptor.getOtcBacklogExpirationInterval();
if (otc_backlog_expiration_interval_in_ms != Config.otc_backlog_expiration_interval_ms_default)
logger.info("OutboundTcpConnection backlog expiration interval set to to {}ms", otc_backlog_expiration_interval_in_ms);
}
private static final MessageOut<?> CLOSE_SENTINEL = new MessageOut(MessagingService.Verb.INTERNAL_RESPONSE);
private volatile boolean isStopped = false;
private static final int OPEN_RETRY_DELAY = 100; // ms between retries
public static final int WAIT_FOR_VERSION_MAX_TIME = 5000;
private static final int NO_VERSION = Integer.MIN_VALUE;
static final int LZ4_HASH_SEED = 0x9747b28c;
private final BlockingQueue<QueuedMessage> backlog = new LinkedBlockingQueue<>();
private static final String BACKLOG_PURGE_SIZE_PROPERTY = PREFIX + "otc_backlog_purge_size";
@VisibleForTesting
static final int BACKLOG_PURGE_SIZE = Integer.getInteger(BACKLOG_PURGE_SIZE_PROPERTY, 1024);
private final AtomicBoolean backlogExpirationActive = new AtomicBoolean(false);
private volatile long backlogNextExpirationTime;
private final OutboundTcpConnectionPool poolReference;
private final CoalescingStrategy cs;
private DataOutputStreamPlus out;
private Socket socket;
private volatile long completed;
private final AtomicLong dropped = new AtomicLong();
private volatile int currentMsgBufferCount = 0;
private volatile int targetVersion;
public OutboundTcpConnection(OutboundTcpConnectionPool pool, String name)
{
super("MessagingService-Outgoing-" + pool.endPoint() + "-" + name);
this.poolReference = pool;
cs = newCoalescingStrategy(pool.endPoint().getHostAddress());
// We want to use the most precise version we know because while there is version detection on connect(),
// the target version might be accessed by the pool (in getConnection()) before we actually connect (as we
// connect when the first message is submitted). Note however that the only case where we'll connect
// without knowing the true version of a node is if that node is a seed (otherwise, we can't know a node
// unless it has been gossiped to us or it has connected to us and in both case this sets the version) and
// in that case we won't rely on that targetVersion before we're actually connected and so the version
// detection in connect() will do its job.
targetVersion = MessagingService.instance().getVersion(pool.endPoint());
}
private static boolean isLocalDC(InetAddress targetHost)
{
String remoteDC = DatabaseDescriptor.getEndpointSnitch().getDatacenter(targetHost);
String localDC = DatabaseDescriptor.getEndpointSnitch().getDatacenter(FBUtilities.getBroadcastAddress());
return remoteDC.equals(localDC);
}
public void enqueue(MessageOut<?> message, int id)
{
long nanoTime = System.nanoTime();
expireMessages(nanoTime);
try
{
backlog.put(new QueuedMessage(message, id, nanoTime));
}
catch (InterruptedException e)
{
throw new AssertionError(e);
}
}
/**
* This is a helper method for unit testing. Disclaimer: Do not use this method outside unit tests, as
* this method is iterating the queue which can be an expensive operation (CPU time, queue locking).
*
* @return true, if the queue contains at least one expired element
*/
@VisibleForTesting // (otherwise = VisibleForTesting.NONE)
boolean backlogContainsExpiredMessages(long nowNanos)
{
return backlog.stream().anyMatch(entry -> entry.isTimedOut(nowNanos));
}
void closeSocket(boolean destroyThread)
{
logger.debug("Enqueuing socket close for {}", poolReference.endPoint());
isStopped = destroyThread; // Exit loop to stop the thread
backlog.clear();
// in the "destroyThread = true" case, enqueuing the sentinel is important mostly to unblock the backlog.take()
// (via the CoalescingStrategy) in case there's a data race between this method enqueuing the sentinel
// and run() clearing the backlog on connection failure.
enqueue(CLOSE_SENTINEL, -1);
}
void softCloseSocket()
{
enqueue(CLOSE_SENTINEL, -1);
}
public int getTargetVersion()
{
return targetVersion;
}
public void run()
{
final int drainedMessageSize = MAX_COALESCED_MESSAGES;
// keeping list (batch) size small for now; that way we don't have an unbounded array (that we never resize)
final List<QueuedMessage> drainedMessages = new ArrayList<>(drainedMessageSize);
outer:
while (!isStopped)
{
try
{
cs.coalesce(backlog, drainedMessages, drainedMessageSize);
}
catch (InterruptedException e)
{
throw new AssertionError(e);
}
int count = currentMsgBufferCount = drainedMessages.size();
//The timestamp of the first message has already been provided to the coalescing strategy
//so skip logging it.
inner:
for (QueuedMessage qm : drainedMessages)
{
try
{
MessageOut<?> m = qm.message;
if (m == CLOSE_SENTINEL)
{
disconnect();
if (isStopped)
break outer;
continue;
}
if (qm.isTimedOut(System.nanoTime()))
dropped.incrementAndGet();
else if (socket != null || connect())
writeConnected(qm, count == 1 && backlog.isEmpty());
else
{
// Not connected! Clear out the queue, else gossip messages back up. Update dropped
// statistics accordingly. Hint: The statistics may be slightly too low, if messages
// are added between the calls of backlog.size() and backlog.clear()
dropped.addAndGet(backlog.size());
backlog.clear();
currentMsgBufferCount = 0;
break inner;
}
}
catch (Exception e)
{
JVMStabilityInspector.inspectThrowable(e);
// really shouldn't get here, as exception handling in writeConnected() is reasonably robust
// but we want to catch anything bad we don't drop the messages in the current batch
logger.error("error processing a message intended for {}", poolReference.endPoint(), e);
}
currentMsgBufferCount = --count;
}
// Update dropped statistics by the number of unprocessed drainedMessages
dropped.addAndGet(currentMsgBufferCount);
drainedMessages.clear();
}
}
public int getPendingMessages()
{
return backlog.size() + currentMsgBufferCount;
}
public long getCompletedMesssages()
{
return completed;
}
public long getDroppedMessages()
{
return dropped.get();
}
private boolean shouldCompressConnection()
{
// assumes version >= 1.2
return DatabaseDescriptor.internodeCompression() == Config.InternodeCompression.all
|| (DatabaseDescriptor.internodeCompression() == Config.InternodeCompression.dc && !isLocalDC(poolReference.endPoint()));
}
private void writeConnected(QueuedMessage qm, boolean flush)
{
try
{
byte[] sessionBytes = qm.message.parameters.get(Tracing.TRACE_HEADER);
if (sessionBytes != null)
{
UUID sessionId = UUIDGen.getUUID(ByteBuffer.wrap(sessionBytes));
TraceState state = Tracing.instance.get(sessionId);
String message = String.format("Sending %s message to %s", qm.message.verb, poolReference.endPoint());
// session may have already finished; see CASSANDRA-5668
if (state == null)
{
byte[] traceTypeBytes = qm.message.parameters.get(Tracing.TRACE_TYPE);
Tracing.TraceType traceType = traceTypeBytes == null ? Tracing.TraceType.QUERY : Tracing.TraceType.deserialize(traceTypeBytes[0]);
Tracing.instance.trace(ByteBuffer.wrap(sessionBytes), message, traceType.getTTL());
}
else
{
state.trace(message);
if (qm.message.verb == MessagingService.Verb.REQUEST_RESPONSE)
Tracing.instance.doneWithNonLocalSession(state);
}
}
long timestampMillis = NanoTimeToCurrentTimeMillis.convert(qm.timestampNanos);
writeInternal(qm.message, qm.id, timestampMillis);
completed++;
if (flush)
out.flush();
}
catch (Throwable e)
{
JVMStabilityInspector.inspectThrowable(e);
disconnect();
if (e instanceof IOException || e.getCause() instanceof IOException)
{
logger.debug("Error writing to {}", poolReference.endPoint(), e);
// If we haven't retried this message yet, put it back on the queue to retry after re-connecting.
// See CASSANDRA-5393 and CASSANDRA-12192.
if (qm.shouldRetry())
{
try
{
backlog.put(new RetriedQueuedMessage(qm));
}
catch (InterruptedException e1)
{
throw new AssertionError(e1);
}
}
}
else
{
// Non IO exceptions are likely a programming error so let's not silence them
logger.error("error writing to {}", poolReference.endPoint(), e);
}
}
}
private void writeInternal(MessageOut<?> message, int id, long timestamp) throws IOException
{
out.writeInt(MessagingService.PROTOCOL_MAGIC);
if (targetVersion < MessagingService.VERSION_20)
out.writeUTF(String.valueOf(id));
else
out.writeInt(id);
// int cast cuts off the high-order half of the timestamp, which we can assume remains
// the same between now and when the recipient reconstructs it.
out.writeInt((int) timestamp);
message.serialize(out, targetVersion);
}
private static void writeHeader(DataOutput out, int version, boolean compressionEnabled) throws IOException
{
// 2 bits: unused. used to be "serializer type," which was always Binary
// 1 bit: compression
// 1 bit: streaming mode
// 3 bits: unused
// 8 bits: version
// 15 bits: unused
int header = 0;
if (compressionEnabled)
header |= 4;
header |= (version << 8);
out.writeInt(header);
}
private void disconnect()
{
if (socket != null)
{
try
{
socket.close();
logger.debug("Socket to {} closed", poolReference.endPoint());
}
catch (IOException e)
{
logger.debug("Exception closing connection to {}", poolReference.endPoint(), e);
}
out = null;
socket = null;
}
}
@SuppressWarnings("resource")
private boolean connect()
{
logger.debug("Attempting to connect to {}", poolReference.endPoint());
long start = System.nanoTime();
long timeout = TimeUnit.MILLISECONDS.toNanos(DatabaseDescriptor.getRpcTimeout());
while (System.nanoTime() - start < timeout && !isStopped)
{
targetVersion = MessagingService.instance().getVersion(poolReference.endPoint());
try
{
socket = poolReference.newSocket();
socket.setKeepAlive(true);
if (isLocalDC(poolReference.endPoint()))
{
socket.setTcpNoDelay(INTRADC_TCP_NODELAY);
}
else
{
socket.setTcpNoDelay(DatabaseDescriptor.getInterDCTcpNoDelay());
}
if (DatabaseDescriptor.getInternodeSendBufferSize() > 0)
{
try
{
socket.setSendBufferSize(DatabaseDescriptor.getInternodeSendBufferSize());
}
catch (SocketException se)
{
logger.warn("Failed to set send buffer size on internode socket.", se);
}
}
// SocketChannel may be null when using SSL
WritableByteChannel ch = socket.getChannel();
out = new BufferedDataOutputStreamPlus(ch != null ? ch : Channels.newChannel(socket.getOutputStream()), BUFFER_SIZE);
out.writeInt(MessagingService.PROTOCOL_MAGIC);
writeHeader(out, targetVersion, shouldCompressConnection());
out.flush();
DataInputStream in = new DataInputStream(socket.getInputStream());
int maxTargetVersion = handshakeVersion(in);
if (maxTargetVersion == NO_VERSION)
{
// no version is returned, so disconnect an try again: we will either get
// a different target version (targetVersion < MessagingService.VERSION_12)
// or if the same version the handshake will finally succeed
logger.trace("Target max version is {}; no version information yet, will retry", maxTargetVersion);
disconnect();
continue;
}
else
{
MessagingService.instance().setVersion(poolReference.endPoint(), maxTargetVersion);
}
if (targetVersion > maxTargetVersion)
{
logger.trace("Target max version is {}; will reconnect with that version", maxTargetVersion);
try
{
if (DatabaseDescriptor.getSeeds().contains(poolReference.endPoint()))
logger.warn("Seed gossip version is {}; will not connect with that version", maxTargetVersion);
}
catch (Throwable e)
{
// If invalid yaml has been added to the config since startup, getSeeds() will throw an AssertionError
// Additionally, third party seed providers may throw exceptions if network is flakey
// Regardless of what's thrown, we must catch it, disconnect, and try again
JVMStabilityInspector.inspectThrowable(e);
logger.warn("Configuration error prevented outbound connection: {}", e.getLocalizedMessage());
}
finally
{
disconnect();
return false;
}
}
if (targetVersion < maxTargetVersion && targetVersion < MessagingService.current_version)
{
logger.trace("Detected higher max version {} (using {}); will reconnect when queued messages are done",
maxTargetVersion, targetVersion);
softCloseSocket();
}
out.writeInt(MessagingService.current_version);
CompactEndpointSerializationHelper.serialize(FBUtilities.getBroadcastAddress(), out);
if (shouldCompressConnection())
{
out.flush();
logger.trace("Upgrading OutputStream to {} to be compressed", poolReference.endPoint());
if (targetVersion < MessagingService.VERSION_21)
{
// Snappy is buffered, so no need for extra buffering output stream
out = new WrappedDataOutputStreamPlus(new SnappyOutputStream(socket.getOutputStream()));
}
else
{
// TODO: custom LZ4 OS that supports BB write methods
LZ4Compressor compressor = LZ4Factory.fastestInstance().fastCompressor();
Checksum checksum = XXHashFactory.fastestInstance().newStreamingHash32(LZ4_HASH_SEED).asChecksum();
out = new WrappedDataOutputStreamPlus(new LZ4BlockOutputStream(socket.getOutputStream(),
1 << 14, // 16k block size
compressor,
checksum,
true)); // no async flushing
}
}
logger.debug("Done connecting to {}", poolReference.endPoint());
return true;
}
catch (SSLHandshakeException e)
{
logger.error("SSL handshake error for outbound connection to " + socket, e);
disconnect();
// SSL errors won't be recoverable within timeout period so we'll just abort
return false;
}
catch (IOException e)
{
disconnect();
logger.debug("Unable to connect to {}", poolReference.endPoint(), e);
Uninterruptibles.sleepUninterruptibly(OPEN_RETRY_DELAY, TimeUnit.MILLISECONDS);
}
}
return false;
}
private int handshakeVersion(final DataInputStream inputStream)
{
final AtomicInteger version = new AtomicInteger(NO_VERSION);
final CountDownLatch versionLatch = new CountDownLatch(1);
NamedThreadFactory.createThread(() ->
{
try
{
logger.info("Handshaking version with {}", poolReference.endPoint());
version.set(inputStream.readInt());
}
catch (IOException ex)
{
final String msg = "Cannot handshake version with " + poolReference.endPoint();
if (logger.isTraceEnabled())
logger.trace(msg, ex);
else
logger.info(msg);
}
finally
{
//unblock the waiting thread on either success or fail
versionLatch.countDown();
}
}, "HANDSHAKE-" + poolReference.endPoint()).start();
try
{
versionLatch.await(WAIT_FOR_VERSION_MAX_TIME, TimeUnit.MILLISECONDS);
}
catch (InterruptedException ex)
{
throw new AssertionError(ex);
}
return version.get();
}
/**
* Expire elements from the queue if the queue is pretty full and expiration is not already in progress.
* This method will only remove droppable expired entries. If no such element exists, nothing is removed from the queue.
*
* @param timestampNanos The current time as from System.nanoTime()
*/
@VisibleForTesting
void expireMessages(long timestampNanos)
{
if (backlog.size() <= BACKLOG_PURGE_SIZE)
return; // Plenty of space
if (backlogNextExpirationTime - timestampNanos > 0)
return; // Expiration is not due.
/**
* Expiration is an expensive process. Iterating the queue locks the queue for both writes and
* reads during iter.next() and iter.remove(). Thus letting only a single Thread do expiration.
*/
if (backlogExpirationActive.compareAndSet(false, true))
{
try
{
Iterator<QueuedMessage> iter = backlog.iterator();
while (iter.hasNext())
{
QueuedMessage qm = iter.next();
if (!qm.droppable)
continue;
if (!qm.isTimedOut(timestampNanos))
continue;
iter.remove();
dropped.incrementAndGet();
}
if (logger.isTraceEnabled())
{
long duration = TimeUnit.NANOSECONDS.toMicros(System.nanoTime() - timestampNanos);
logger.trace("Expiration of {} took {}μs", getName(), duration);
}
}
finally
{
long backlogExpirationIntervalNanos = TimeUnit.MILLISECONDS.toNanos(DatabaseDescriptor.getOtcBacklogExpirationInterval());
backlogNextExpirationTime = timestampNanos + backlogExpirationIntervalNanos;
backlogExpirationActive.set(false);
}
}
}
/** messages that have not been retried yet */
private static class QueuedMessage implements Coalescable
{
final MessageOut<?> message;
final int id;
final long timestampNanos;
final boolean droppable;
QueuedMessage(MessageOut<?> message, int id, long timestampNanos)
{
this.message = message;
this.id = id;
this.timestampNanos = timestampNanos;
this.droppable = MessagingService.DROPPABLE_VERBS.contains(message.verb);
}
/** don't drop a non-droppable message just because it's timestamp is expired */
boolean isTimedOut(long nowNanos)
{
long messageTimeoutNanos = TimeUnit.MILLISECONDS.toNanos(message.getTimeout());
return droppable && nowNanos - timestampNanos > messageTimeoutNanos;
}
boolean shouldRetry()
{
// retry all messages once
return true;
}
public long timestampNanos()
{
return timestampNanos;
}
}
private static class RetriedQueuedMessage extends QueuedMessage
{
RetriedQueuedMessage(QueuedMessage msg)
{
super(msg.message, msg.id, msg.timestampNanos);
}
boolean shouldRetry()
{
return false;
}
}
}