本帖最后由 坎蒂丝_Swan 于 2015-1-12 14:20 编辑
问题导读
1.文件sink使用时,如何配置?
2.文件sink又有哪些依赖包?
应用场景:需要实时收集多台服务器的nginx日志到一台机器。收集完成结果存放需要按天生成文件夹,按每5分钟生成文件,比如2012年12月29日12点26分的日志,需要放到/data/log/20121229/log-1225-对应的文件中。自己实现了类似flume-og和flume-ng的hdfs-sink的文件sink。
使用的时候配置如下:
agent.sources = source
agent.channels = channel
agent.sinks = sink
agent.sources.source.type = avro
agent.sources.source.bind = 192.168.0.100
agent.sources.source.port = 44444
agent.sources.source.channels = channel
agent.sinks.sink.type = org.apache.flume.sink.FileSink
agent.sinks.sink.file.path = /data/log/%{dayStr}
agent.sinks.sink.file.filePrefix = log-%{hourStr}%{minStr}-
agent.sinks.sink.file.txnEventMax = 10000
agent.sinks.sink.file.maxOpenFiles = 5
agent.sinks.sink.channel = channel
agent.channels.channel.type = memory
agent.channels.channel.capacity = 100000
agent.channels.channel.transactionCapacity = 100000
agent.channels.channel.keep-alive = 60 复制代码
依赖的jar如下:
jakarta-oro-2.0.1.jar
flume-ng-core-1.3.0-SNAPSHOT.jar
flume-ng-sdk-1.3.0-SNAPSHOT.jar
flume-ng-configuration-1.3.0-SNAPSHOT.jar
slf4j-log4j12-1.6.1.jar
slf4j-api-1.6.1.jar
guava-10.0.1.jar
代码如下:
FileSink.java
package org.apache.flume.sink;
import java.io.IOException;
import java.util.Calendar;
import java.util.List;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import org.apache.flume.Channel;
import org.apache.flume.Context;
import org.apache.flume.Event;
import org.apache.flume.EventDeliveryException;
import org.apache.flume.Transaction;
import org.apache.flume.conf.Configurable;
import org.apache.flume.formatter.output.BucketPath;
import org.apache.flume.instrumentation.SinkCounter;
import org.apache.flume.serialization.EventSerializer;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.google.common.base.Preconditions;
import com.google.common.collect.Lists;
import com.google.common.util.concurrent.ThreadFactoryBuilder;
public class FileSink extends AbstractSink implements Configurable {
private static final Logger logger = LoggerFactory
. getLogger(FileSink .class );
private String path ;
private static final String defaultFileName = "FlumeData";
private static final int defaultMaxOpenFiles = 50;
/**
* Default length of time we wait for blocking BucketWriter calls before
* timing out the operation. Intended to prevent server hangs.
*/
private long txnEventMax ;
private FileWriterLinkedHashMap sfWriters ;
private String serializerType ;
private Context serializerContext ;
private boolean needRounding = false;
private int roundUnit = Calendar.SECOND;
private int roundValue = 1;
private SinkCounter sinkCounter ;
private int maxOpenFiles ;
private ScheduledExecutorService timedRollerPool ;
private long rollInterval ;
@Override
public void configure(Context context) {
String directory = Preconditions. checkNotNull(
context.getString( "file.path"), "file.path is required");
String fileName = context.getString( "file.filePrefix", defaultFileName);
this.path = directory + "/" + fileName;
maxOpenFiles = context.getInteger("file.maxOpenFiles" ,
defaultMaxOpenFiles);
serializerType = context.getString("sink.serializer" , "TEXT" );
serializerContext = new Context(
context.getSubProperties(EventSerializer. CTX_PREFIX));
txnEventMax = context.getLong("file.txnEventMax" , 1l);
if (sinkCounter == null) {
sinkCounter = new SinkCounter(getName());
}
rollInterval = context.getLong("file.rollInterval" , 30l);
String rollerName = "hdfs-" + getName() + "-roll-timer-%d" ;
timedRollerPool = Executors.newScheduledThreadPool( maxOpenFiles,
new ThreadFactoryBuilder().setNameFormat(rollerName).build());
}
@Override
public Status process() throws EventDeliveryException {
Channel channel = getChannel();
Transaction transaction = channel.getTransaction();
List<BucketFileWriter> writers = Lists. newArrayList();
transaction.begin();
try {
Event event = null;
int txnEventCount = 0;
for (txnEventCount = 0; txnEventCount < txnEventMax; txnEventCount++) {
event = channel.take();
if (event == null) {
break;
}
// reconstruct the path name by substituting place holders
String realPath = BucketPath
. escapeString(path, event.getHeaders(), needRounding,
roundUnit, roundValue );
BucketFileWriter bucketFileWriter = sfWriters.get(realPath);
// we haven't seen this file yet, so open it and cache the
// handle
if (bucketFileWriter == null) {
bucketFileWriter = new BucketFileWriter();
bucketFileWriter.open(realPath, serializerType,
serializerContext, rollInterval , timedRollerPool,
sfWriters);
sfWriters.put(realPath, bucketFileWriter);
}
// track the buckets getting written in this transaction
if (!writers.contains(bucketFileWriter)) {
writers.add(bucketFileWriter);
}
// Write the data to File
bucketFileWriter.append(event);
}
if (txnEventCount == 0) {
sinkCounter.incrementBatchEmptyCount();
} else if (txnEventCount == txnEventMax) {
sinkCounter.incrementBatchCompleteCount();
} else {
sinkCounter.incrementBatchUnderflowCount();
}
// flush all pending buckets before committing the transaction
for (BucketFileWriter bucketFileWriter : writers) {
if (!bucketFileWriter.isBatchComplete()) {
flush(bucketFileWriter);
}
}
transaction.commit();
if (txnEventCount > 0) {
sinkCounter.addToEventDrainSuccessCount(txnEventCount);
}
if (event == null) {
return Status.BACKOFF ;
}
return Status.READY ;
} catch (IOException eIO) {
transaction.rollback();
logger.warn("File IO error" , eIO);
return Status.BACKOFF ;
} catch (Throwable th) {
transaction.rollback();
logger.error("process failed" , th);
if (th instanceof Error) {
throw (Error) th;
} else {
throw new EventDeliveryException(th);
}
} finally {
transaction.close();
}
}
private void flush(BucketFileWriter bucketFileWriter) throws IOException {
bucketFileWriter.flush();
}
@Override
public synchronized void start() {
super.start();
this.sfWriters = new FileWriterLinkedHashMap(maxOpenFiles);
sinkCounter.start();
}
} 复制代码
BucketFileWriter.java
package org.apache.flume.sink;
import java.io.BufferedOutputStream;
import java.io.File;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.OutputStream;
import java.util.concurrent.Callable;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicLong;
import org.apache.flume.Context;
import org.apache.flume.Event;
import org.apache.flume.serialization.EventSerializer;
import org.apache.flume.serialization.EventSerializerFactory;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
public class BucketFileWriter {
private static final Logger logger = LoggerFactory
.getLogger(BucketFileWriter.class);
private static final String IN_USE_EXT = ".tmp";
/**
* This lock ensures that only one thread can open a file at a time.
*/
private final AtomicLong fileExtensionCounter;
private OutputStream outputStream;
private EventSerializer serializer;
private String filePath;
/**
* Close the file handle and rename the temp file to the permanent filename.
* Safe to call multiple times. Logs HDFSWriter.close() exceptions.
*
* @throws IOException
* On failure to rename if temp file exists.
*/
public BucketFileWriter() {
fileExtensionCounter = new AtomicLong(System.currentTimeMillis());
}
public void open(final String filePath, String serializerType,
Context serializerContext, final long rollInterval,
final ScheduledExecutorService timedRollerPool,
final FileWriterLinkedHashMap sfWriters) throws IOException {
this.filePath = filePath;
File file = new File(filePath + fileExtensionCounter + IN_USE_EXT);
file.getParentFile().mkdirs();
outputStream = new BufferedOutputStream(new FileOutputStream(file));
logger.info("filename = " + file.getAbsolutePath());
serializer = EventSerializerFactory.getInstance(serializerType,
serializerContext, outputStream);
serializer.afterCreate();
if (rollInterval > 0) {
Callable<Void> action = new Callable<Void>() {
@Override
public Void call() throws Exception {
logger.debug(
"Rolling file ({}): Roll scheduled after {} sec elapsed.",
filePath + fileExtensionCounter + IN_USE_EXT,
rollInterval);
if (sfWriters.containsKey(filePath)) {
sfWriters.remove(filePath);
}
close();
return null;
}
};
timedRollerPool.schedule(action, rollInterval, TimeUnit.SECONDS);
}
}
public void append(Event event) throws IOException {
serializer.write(event);
}
public boolean isBatchComplete() {
return true;
}
public void flush() throws IOException {
serializer.flush();
outputStream.flush();
}
/**
* Rename bucketPath file from .tmp to permanent location.
*/
private void renameBucket() {
File srcPath = new File(filePath + fileExtensionCounter + IN_USE_EXT);
File dstPath = new File(filePath + fileExtensionCounter);
if (srcPath.exists()) {
srcPath.renameTo(dstPath);
logger.info("Renaming " + srcPath + " to " + dstPath);
}
}
public synchronized void close() throws IOException, InterruptedException {
if (outputStream != null) {
outputStream.flush();
outputStream.close();
}
renameBucket();
}
} 复制代码
FileWriterLinkedHashMap.java
package org.apache.flume.sink;
import java.io.IOException;
import java.util.LinkedHashMap;
import java.util.Map.Entry;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
public class FileWriterLinkedHashMap extends
LinkedHashMap<String, BucketFileWriter> {
private static final Logger logger = LoggerFactory
.getLogger(FileWriterLinkedHashMap.class);
private static final long serialVersionUID = -7860596835613215998L;
private final int maxOpenFiles;
public FileWriterLinkedHashMap(int maxOpenFiles) {
super(16, 0.75f, true); // stock initial capacity/load, access
this.maxOpenFiles = maxOpenFiles;
}
@Override
protected boolean removeEldestEntry(Entry<String, BucketFileWriter> eldest) {
if (size() > maxOpenFiles) {
// If we have more that max open files, then close the last one
// and
// return true
try {
eldest.getValue().close();
} catch (IOException e) {
logger.warn(eldest.getKey().toString(), e);
} catch (InterruptedException e) {
logger.warn(eldest.getKey().toString(), e);
Thread.currentThread().interrupt();
}
return true;
} else {
return false;
}
}
} 复制代码