/*
* Fetch a file that is in a Hadoop file system. Return a local File.
* Interruptible.
*/
private File hdfsFetch(Path fromPath, Reporter reporter)
throws IOException, InterruptedException {
UUID uniqueId = UUID.randomUUID();
File toFile = new File(tempDir, uniqueId.toString() + "/" + fromPath.getName());
File toDir = new File(toFile.getParent());
if (toDir.exists()) {
FileUtils.deleteDirectory(toDir);
}
toDir.mkdirs();
Path toPath = new Path(toFile.getCanonicalPath());
FileSystem fS = fromPath.getFileSystem(hadoopConf);
FileSystem tofS = FileSystem.getLocal(hadoopConf);
Throttler throttler = new Throttler((double) bytesPerSecThrottle);
try {
for (FileStatus fStatus : fS.globStatus(fromPath)) {
log.info("Copying " + fStatus.getPath() + " to " + toPath);
long bytesSoFar = 0;
FSDataInputStream iS = fS.open(fStatus.getPath());
FSDataOutputStream oS = tofS.create(toPath);
byte[] buffer = new byte[downloadBufferSize];
int nRead;
while ((nRead = iS.read(buffer, 0, buffer.length)) != -1) {
// Needed to being able to be interrupted at any moment.
if (Thread.interrupted()) {
iS.close();
oS.close();
cleanDirNoExceptions(toDir);
throw new InterruptedException();
}
bytesSoFar += nRead;
oS.write(buffer, 0, nRead);
throttler.incrementAndThrottle(nRead);
if (bytesSoFar >= bytesToReportProgress) {
reporter.progress(bytesSoFar);
bytesSoFar = 0l;
}
}
if (reporter != null) {
reporter.progress(bytesSoFar);
}
oS.close();
iS.close();
}
return toDir;
} catch (ClosedByInterruptException e) {
// This can be thrown by the method read.
cleanDirNoExceptions(toDir);
throw new InterruptedIOException();
}
}
private void copyFile(File sourceFile, File destFile, Reporter reporter) throws IOException {
if (!destFile.exists()) {
destFile.createNewFile();
}
FileChannel source = null;
FileChannel destination = null;
Throttler throttler = new Throttler((double) bytesPerSecThrottle);
FileInputStream iS = null;
FileOutputStream oS = null;
try {
iS = new FileInputStream(sourceFile);
oS = new FileOutputStream(destFile);
source = iS.getChannel();
destination = oS.getChannel();
long bytesSoFar = 0;
long reportingBytesSoFar = 0;
long size = source.size();
int transferred = 0;
while (bytesSoFar < size) {
// Casting to int here is safe since we will transfer at most "downloadBufferSize" bytes.
// This is done on purpose for being able to implement Throttling.
transferred = (int) destination.transferFrom(source, bytesSoFar, downloadBufferSize);
bytesSoFar += transferred;
reportingBytesSoFar += transferred;
throttler.incrementAndThrottle(transferred);
if (reportingBytesSoFar >= bytesToReportProgress) {
reporter.progress(reportingBytesSoFar);
reportingBytesSoFar = 0l;
}
}
if (reporter != null) {
reporter.progress(reportingBytesSoFar);
}
} finally {
if (iS != null) {
iS.close();
}
if (oS != null) {
oS.close();
}
if (source != null) {
source.close();
}
if (destination != null) {
destination.close();
}
}
}
public synchronized void flush() throws IOException {
LOG.info("Starting flush of map output");
synchronized (spillLock) {
while (kvstart != kvend) {
try {
reporter.progress();
spillLock.wait();
} catch (InterruptedException e) {
throw (IOException)
new IOException("Buffer interrupted while waiting for the writer").initCause(e);
}
}
}
if (sortSpillException != null) {
throw (IOException) new IOException("Spill failed").initCause(sortSpillException);
}
if (kvend != kvindex) {
LOG.info("bufstart = " + bufstart + "; bufend = " + bufmark + "; bufvoid = " + bufvoid);
LOG.info(
"kvstart = " + kvstart + "; kvend = " + kvindex + "; length = " + kvoffsets.length);
kvend = kvindex;
bufend = bufmark;
sortAndSpill();
}
// release sort buffer before the merge
kvbuffer = null;
mergeParts();
}
/*
* Fetch a file that is in a Hadoop file system. Return a local File.
*/
private File hdfsFetch(Path fromPath, Reporter reporter) throws IOException {
File toFile = new File(tempDir, fromPath.toUri().getPath());
File toDir = new File(toFile.getParent());
if (toDir.exists()) {
FileUtils.deleteDirectory(toDir);
}
toDir.mkdirs();
Path toPath = new Path(toFile.getCanonicalPath());
FileSystem fS = fromPath.getFileSystem(hadoopConf);
FileSystem tofS = FileSystem.getLocal(hadoopConf);
Throttler throttler = new Throttler((double) bytesPerSecThrottle);
for (FileStatus fStatus : fS.globStatus(fromPath)) {
log.info("Copying " + fStatus.getPath() + " to " + toPath);
long bytesSoFar = 0;
FSDataInputStream iS = fS.open(fStatus.getPath());
FSDataOutputStream oS = tofS.create(toPath);
byte[] buffer = new byte[downloadBufferSize];
int nRead;
while ((nRead = iS.read(buffer, 0, buffer.length)) != -1) {
bytesSoFar += nRead;
oS.write(buffer, 0, nRead);
throttler.incrementAndThrottle(nRead);
if (bytesSoFar >= bytesToReportProgress) {
reporter.progress(bytesSoFar);
bytesSoFar = 0l;
}
}
if (reporter != null) {
reporter.progress(bytesSoFar);
}
oS.close();
iS.close();
}
return toDir;
}
protected void waitForOpenSlot(int maxProcessesOnNode, Reporter reporter)
throws IOException, InterruptedException {
while (true) {
// sleep for a random length of time between 0 and 60 seconds
long sleepTime = (long) (Math.random() * 1000 * 60);
logger.info("sleeping for " + sleepTime);
Thread.sleep(sleepTime);
int numRunningMappers = getNumRunningMappers();
logger.info("num running mappers: " + numRunningMappers);
if (numRunningMappers < maxProcessesOnNode) return;
reporter.progress();
}
}
public void map(
LongWritable key, Text value, OutputCollector<Text, Text> output, Reporter reporter)
throws IOException {
if (this.output == null) {
this.output = output;
}
if (this.reporter == null) {
this.reporter = reporter;
}
String line = value.toString();
String[] reads = line.split("\n");
splitRead(key, reads[0], s1FileWriter);
splitRead(key, reads[1], getRead2FileWriter());
reporter.progress();
}
@SuppressWarnings("unchecked")
private void combineAndSpill(RawKeyValueIterator kvIter, Counters.Counter inCounter)
throws IOException {
Reducer combiner = (Reducer) ReflectionUtils.newInstance(combinerClass, job);
try {
CombineValuesIterator values =
new CombineValuesIterator(
kvIter, comparator, keyClass, valClass, job, reporter, inCounter);
while (values.more()) {
combiner.reduce(values.getKey(), values, combineCollector, reporter);
values.nextKey();
// indicate we're making progress
reporter.progress();
}
} finally {
combiner.close();
}
}
@SuppressWarnings("unchecked")
public synchronized void collect(K key, V value) throws IOException {
reporter.progress();
if (key.getClass() != keyClass) {
throw new IOException(
"Type mismatch in key from map: expected "
+ keyClass.getName()
+ ", recieved "
+ key.getClass().getName());
}
if (value.getClass() != valClass) {
throw new IOException(
"Type mismatch in value from map: expected "
+ valClass.getName()
+ ", recieved "
+ value.getClass().getName());
}
if (sortSpillException != null) {
throw (IOException) new IOException("Spill failed").initCause(sortSpillException);
}
try {
// serialize key bytes into buffer
int keystart = bufindex;
keySerializer.serialize(key);
if (bufindex < keystart) {
// wrapped the key; reset required
bb.reset();
keystart = 0;
}
// serialize value bytes into buffer
int valstart = bufindex;
valSerializer.serialize(value);
int valend = bb.markRecord();
mapOutputByteCounter.increment(
valend >= keystart ? valend - keystart : (bufvoid - keystart) + valend);
if (keystart == bufindex) {
// if emitted records make no writes, it's possible to wrap
// accounting space without notice
bb.write(new byte[0], 0, 0);
}
int partition = partitioner.getPartition(key, value, partitions);
if (partition < 0 || partition >= partitions) {
throw new IOException("Illegal partition for " + key + " (" + partition + ")");
}
mapOutputRecordCounter.increment(1);
// update accounting info
int ind = kvindex * ACCTSIZE;
kvoffsets[kvindex] = ind;
kvindices[ind + PARTITION] = partition;
kvindices[ind + KEYSTART] = keystart;
kvindices[ind + VALSTART] = valstart;
kvindex = (kvindex + 1) % kvoffsets.length;
} catch (MapBufferTooSmallException e) {
LOG.info("Record too large for in-memory buffer: " + e.getMessage());
spillSingleRecord(key, value);
mapOutputRecordCounter.increment(1);
return;
}
}
public void collect(K key, V value) throws IOException {
reporter.progress();
out.write(key, value);
mapOutputRecordCounter.increment(1);
}
/*
* Fetch a file that is in a S3 file system. Return a local File. It accepts "s3://" and "s3n://" prefixes.
*/
private File s3Fetch(URI uri, Reporter reporter) throws IOException {
String bucketName = uri.getHost();
String path = uri.getPath();
File destFolder = new File(tempDir, bucketName + "/" + path);
if (destFolder.exists()) {
FileUtils.deleteDirectory(destFolder);
}
destFolder.mkdirs();
Throttler throttler = new Throttler((double) bytesPerSecThrottle);
boolean done = false;
try {
s3Service = new RestS3Service(getCredentials());
if (s3Service.checkBucketStatus(bucketName) != RestS3Service.BUCKET_STATUS__MY_BUCKET) {
throw new IOException("Bucket doesn't exist or is already claimed: " + bucketName);
}
if (path.startsWith("/")) {
path = path.substring(1, path.length());
}
for (S3Object object : s3Service.listObjects(new S3Bucket(bucketName), path, "")) {
long bytesSoFar = 0;
String fileName = path;
if (path.contains("/")) {
fileName = path.substring(path.lastIndexOf("/") + 1, path.length());
}
File fileDest = new File(destFolder, fileName);
log.info("Downloading " + object.getKey() + " to " + fileDest + " ...");
if (fileDest.exists()) {
fileDest.delete();
}
object = s3Service.getObject(new S3Bucket(bucketName), object.getKey());
InputStream iS = object.getDataInputStream();
FileOutputStream writer = new FileOutputStream(fileDest);
byte[] buffer = new byte[downloadBufferSize];
int nRead;
while ((nRead = iS.read(buffer, 0, buffer.length)) != -1) {
bytesSoFar += nRead;
writer.write(buffer, 0, nRead);
throttler.incrementAndThrottle(nRead);
if (bytesSoFar >= bytesToReportProgress) {
reporter.progress(bytesSoFar);
bytesSoFar = 0l;
}
}
if (reporter != null) {
reporter.progress(bytesSoFar);
}
writer.close();
iS.close();
done = true;
}
if (!done) {
throw new IOException("Bucket is empty! " + bucketName + " path: " + path);
}
} catch (S3ServiceException e) {
throw new IOException(e);
}
return destFolder;
}
