/*
* 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.flume.sink;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.Preconditions;
import com.google.common.base.Throwables;
import com.google.common.collect.Lists;
import com.google.common.util.concurrent.ThreadFactoryBuilder;
import org.apache.flume.Channel;
import org.apache.flume.ChannelException;
import org.apache.flume.Context;
import org.apache.flume.Event;
import org.apache.flume.EventDeliveryException;
import org.apache.flume.FlumeException;
import org.apache.flume.Transaction;
import org.apache.flume.api.RpcClient;
import org.apache.flume.api.RpcClientConfigurationConstants;
import org.apache.flume.conf.Configurable;
import org.apache.flume.instrumentation.SinkCounter;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import java.util.List;
import java.util.Map.Entry;
import java.util.Properties;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.locks.Lock;
import java.util.concurrent.locks.ReentrantLock;
/**
* This sink provides the basic RPC functionality for Flume. This sink takes
* several arguments which are used in RPC.
* This sink forms one half of Flume's tiered collection support. Events sent to
* this sink are transported over the network to the hostname / port pair using
* the RPC implementation encapsulated in {@link RpcClient}.
* The destination is an instance of Flume's {@link org.apache.flume.source
* .AvroSource} or {@link org.apache.flume.source.ThriftSource} (based on
* which implementation of this class is used), which
* allows Flume agents to forward to other Flume agents, forming a tiered
* collection infrastructure. Of course, nothing prevents one from using this
* sink to speak to other custom built infrastructure that implements the same
* RPC protocol.
* </p>
* <p>
* Events are taken from the configured {@link Channel} in batches of the
* configured <tt>batch-size</tt>. The batch size has no theoretical limits
* although all events in the batch <b>must</b> fit in memory. Generally, larger
* batches are far more efficient, but introduce a slight delay (measured in
* millis) in delivery. The batch behavior is such that underruns (i.e. batches
* smaller than the configured batch size) are possible. This is a compromise
* made to maintain low latency of event delivery. If the channel returns a null
* event, meaning it is empty, the batch is immediately sent, regardless of
* size. Batch underruns are tracked in the metrics. Empty batches do not incur
* an RPC roundtrip.
* </p>
* <p>
* <b>Configuration options</b>
* </p>
* <table>
* <tr>
* <th>Parameter</th>
* <th>Description</th>
* <th>Unit (data type)</th>
* <th>Default</th>
* </tr>
* <tr>
* <td><tt>hostname</tt></td>
* <td>The hostname to which events should be sent.</td>
* <td>Hostname or IP (String)</td>
* <td>none (required)</td>
* </tr>
* <tr>
* <td><tt>port</tt></td>
* <td>The port to which events should be sent on <tt>hostname</tt>.</td>
* <td>TCP port (int)</td>
* <td>none (required)</td>
* </tr>
* <tr>
* <td><tt>batch-size</tt></td>
* <td>The maximum number of events to send per RPC.</td>
* <td>events (int)</td>
* <td>100</td>
* </tr>
* <tr>
* <td><tt>connect-timeout</tt></td>
* <td>Maximum time to wait for the first Avro handshake and RPC request</td>
* <td>milliseconds (long)</td>
* <td>20000</td>
* </tr>
* <tr>
* <td><tt>request-timeout</tt></td>
* <td>Maximum time to wait RPC requests after the first</td>
* <td>milliseconds (long)</td>
* <td>20000</td>
* </tr>
* <tr>
* <td><tt>compression-type</tt></td>
* <td>Select compression type. Default is "none" and the only compression type available is "deflate"</td>
* <td>compression type</td>
* <td>none</td>
* </tr>
* <tr>
* <td><tt>compression-level</tt></td>
* <td>In the case compression type is "deflate" this value can be between 0-9. 0 being no compression and
* 1-9 is compression. The higher the number the better the compression. 6 is the default.</td>
* <td>compression level</td>
* <td>6</td>
* </tr>
* </table>
* <p>
* <b>Metrics</b>
* </p>
* <p>
* TODO
* </p>
*
* <strong>Implementation Notes:</strong> Any implementation of this class
* must override the {@linkplain #initializeRpcClient(Properties)} method.
* This method will be called whenever this sink needs to create a new
* connection to the source.
*/
public abstract class AbstractRpcSink extends AbstractSink
implements Configurable {
private static final Logger logger = LoggerFactory.getLogger
(AbstractRpcSink.class);
private String hostname;
private Integer port;
private RpcClient client;
private Properties clientProps;
private SinkCounter sinkCounter;
private int cxnResetInterval;
private AtomicBoolean resetConnectionFlag;
private final int DEFAULT_CXN_RESET_INTERVAL = 0;
private final ScheduledExecutorService cxnResetExecutor = Executors
.newSingleThreadScheduledExecutor(new ThreadFactoryBuilder()
.setNameFormat("Rpc Sink Reset Thread").build());
@Override
public void configure(Context context) {
clientProps = new Properties();
hostname = context.getString("hostname");
port = context.getInteger("port");
Preconditions.checkState(hostname != null, "No hostname specified");
Preconditions.checkState(port != null, "No port specified");
clientProps.setProperty(RpcClientConfigurationConstants.CONFIG_HOSTS, "h1");
clientProps.setProperty(RpcClientConfigurationConstants.CONFIG_HOSTS_PREFIX +
"h1", hostname + ":" + port);
for (Entry<String, String> entry: context.getParameters().entrySet()) {
clientProps.setProperty(entry.getKey(), entry.getValue());
}
if (sinkCounter == null) {
sinkCounter = new SinkCounter(getName());
}
cxnResetInterval = context.getInteger("reset-connection-interval",
DEFAULT_CXN_RESET_INTERVAL);
if(cxnResetInterval == DEFAULT_CXN_RESET_INTERVAL) {
logger.info("Connection reset is set to " + String.valueOf
(DEFAULT_CXN_RESET_INTERVAL) +". Will not reset connection to next " +
"hop");
}
}
/**
* Returns a new {@linkplain RpcClient} instance configured using the given
* {@linkplain Properties} object. This method is called whenever a new
* connection needs to be created to the next hop.
* @param props
* @return
*/
protected abstract RpcClient initializeRpcClient(Properties props);
/**
* If this function is called successively without calling
* {@see #destroyConnection()}, only the first call has any effect.
* @throws org.apache.flume.FlumeException if an RPC client connection could not be opened
*/
private void createConnection() throws FlumeException {
if (client == null) {
logger.info("Rpc sink {}: Building RpcClient with hostname: {}, " +
"port: {}",
new Object[] { getName(), hostname, port });
try {
resetConnectionFlag = new AtomicBoolean(false);
client = initializeRpcClient(clientProps);
Preconditions.checkNotNull(client, "Rpc Client could not be " +
"initialized. " + getName() + " could not be started");
sinkCounter.incrementConnectionCreatedCount();
if (cxnResetInterval > 0) {
cxnResetExecutor.schedule(new Runnable() {
@Override
public void run() {
resetConnectionFlag.set(true);
}
}, cxnResetInterval, TimeUnit.SECONDS);
}
} catch (Exception ex) {
sinkCounter.incrementConnectionFailedCount();
if (ex instanceof FlumeException) {
throw (FlumeException) ex;
} else {
throw new FlumeException(ex);
}
}
logger.debug("Rpc sink {}: Created RpcClient: {}", getName(), client);
}
}
private void resetConnection() {
try {
destroyConnection();
createConnection();
} catch (Throwable throwable) {
//Don't rethrow, else this runnable won't get scheduled again.
logger.error("Error while trying to expire connection",
throwable);
}
}
private void destroyConnection() {
if (client != null) {
logger.debug("Rpc sink {} closing Rpc client: {}", getName(), client);
try {
client.close();
sinkCounter.incrementConnectionClosedCount();
} catch (FlumeException e) {
sinkCounter.incrementConnectionFailedCount();
logger.error("Rpc sink " + getName() + ": Attempt to close Rpc " +
"client failed. Exception follows.", e);
}
}
client = null;
}
/**
* Ensure the connection exists and is active.
* If the connection is not active, destroy it and recreate it.
*
* @throws org.apache.flume.FlumeException If there are errors closing or opening the RPC
* connection.
*/
private void verifyConnection() throws FlumeException {
if (client == null) {
createConnection();
} else if (!client.isActive()) {
destroyConnection();
createConnection();
}
}
/**
* The start() of RpcSink is more of an optimization that allows connection
* to be created before the process() loop is started. In case it so happens
* that the start failed, the process() loop will itself attempt to reconnect
* as necessary. This is the expected behavior since it is possible that the
* downstream source becomes unavailable in the middle of the process loop
* and the sink will have to retry the connection again.
*/
@Override
public void start() {
logger.info("Starting {}...", this);
sinkCounter.start();
try {
createConnection();
} catch (FlumeException e) {
logger.warn("Unable to create Rpc client using hostname: " + hostname
+ ", port: " + port, e);
/* Try to prevent leaking resources. */
destroyConnection();
}
super.start();
logger.info("Rpc sink {} started.", getName());
}
@Override
public void stop() {
logger.info("Rpc sink {} stopping...", getName());
destroyConnection();
cxnResetExecutor.shutdown();
try {
if (cxnResetExecutor.awaitTermination(5, TimeUnit.SECONDS)) {
cxnResetExecutor.shutdownNow();
}
} catch (Exception ex) {
logger.error("Interrupted while waiting for connection reset executor " +
"to shut down");
}
sinkCounter.stop();
super.stop();
logger.info("Rpc sink {} stopped. Metrics: {}", getName(), sinkCounter);
}
@Override
public String toString() {
return "RpcSink " + getName() + " { host: " + hostname + ", port: " +
port + " }";
}
@Override
public Status process() throws EventDeliveryException {
Status status = Status.READY;
Channel channel = getChannel();
Transaction transaction = channel.getTransaction();
if(resetConnectionFlag.get()) {
resetConnection();
// if the time to reset is long and the timeout is short
// this may cancel the next reset request
// this should however not be an issue
resetConnectionFlag.set(false);
}
try {
transaction.begin();
verifyConnection();
List<Event> batch = Lists.newLinkedList();
for (int i = 0; i < client.getBatchSize(); i++) {
Event event = channel.take();
if (event == null) {
break;
}
batch.add(event);
}
int size = batch.size();
int batchSize = client.getBatchSize();
if (size == 0) {
sinkCounter.incrementBatchEmptyCount();
status = Status.BACKOFF;
} else {
if (size < batchSize) {
sinkCounter.incrementBatchUnderflowCount();
} else {
sinkCounter.incrementBatchCompleteCount();
}
sinkCounter.addToEventDrainAttemptCount(size);
client.appendBatch(batch);
}
transaction.commit();
sinkCounter.addToEventDrainSuccessCount(size);
} catch (Throwable t) {
transaction.rollback();
if (t instanceof Error) {
throw (Error) t;
} else if (t instanceof ChannelException) {
logger.error("Rpc Sink " + getName() + ": Unable to get event from" +
" channel " + channel.getName() + ". Exception follows.", t);
status = Status.BACKOFF;
} else {
destroyConnection();
throw new EventDeliveryException("Failed to send events", t);
}
} finally {
transaction.close();
}
return status;
}
@VisibleForTesting
RpcClient getUnderlyingClient() {
return client;
}
}