protected DirectoryScanner createPartition(int partitionIndex, int partitionCount) {
DirectoryScanner that = new DirectoryScanner();
that.filePatternRegexp = this.filePatternRegexp;
that.regex = this.regex;
that.partitionIndex = partitionIndex;
that.partitionCount = partitionCount;
return that;
}
/** Scans the directory for new files. */
protected void scanDirectory() {
if (System.currentTimeMillis() - scanIntervalMillis >= lastScanMillis) {
Set<Path> newPaths = scanner.scan(fs, filePath, processedFiles);
for (Path newPath : newPaths) {
String newPathString = newPath.toString();
pendingFiles.add(newPathString);
processedFiles.add(newPathString);
localProcessedFileCount.increment();
}
lastScanMillis = System.currentTimeMillis();
}
}
@Override
public Collection<Partition<AbstractFileInputOperator<T>>> definePartitions(
Collection<Partition<AbstractFileInputOperator<T>>> partitions, PartitioningContext context) {
lastRepartition = System.currentTimeMillis();
int totalCount = getNewPartitionCount(partitions, context);
LOG.debug("Computed new partitions: {}", totalCount);
if (totalCount == partitions.size()) {
return partitions;
}
AbstractFileInputOperator<T> tempOperator =
partitions.iterator().next().getPartitionedInstance();
MutableLong tempGlobalNumberOfRetries = tempOperator.globalNumberOfRetries;
MutableLong tempGlobalNumberOfFailures = tempOperator.globalNumberOfRetries;
/*
* Build collective state from all instances of the operator.
*/
Set<String> totalProcessedFiles = Sets.newHashSet();
Set<FailedFile> currentFiles = Sets.newHashSet();
List<DirectoryScanner> oldscanners = Lists.newLinkedList();
List<FailedFile> totalFailedFiles = Lists.newLinkedList();
List<String> totalPendingFiles = Lists.newLinkedList();
Set<Integer> deletedOperators = Sets.newHashSet();
for (Partition<AbstractFileInputOperator<T>> partition : partitions) {
AbstractFileInputOperator<T> oper = partition.getPartitionedInstance();
totalProcessedFiles.addAll(oper.processedFiles);
totalFailedFiles.addAll(oper.failedFiles);
totalPendingFiles.addAll(oper.pendingFiles);
currentFiles.addAll(unfinishedFiles);
tempGlobalNumberOfRetries.add(oper.localNumberOfRetries);
tempGlobalNumberOfFailures.add(oper.localNumberOfFailures);
if (oper.currentFile != null) {
currentFiles.add(new FailedFile(oper.currentFile, oper.offset));
}
oldscanners.add(oper.getScanner());
deletedOperators.add(oper.operatorId);
}
/*
* Create partitions of scanners, scanner's partition method will do state
* transfer for DirectoryScanner objects.
*/
List<DirectoryScanner> scanners = scanner.partition(totalCount, oldscanners);
Kryo kryo = new Kryo();
Collection<Partition<AbstractFileInputOperator<T>>> newPartitions =
Lists.newArrayListWithExpectedSize(totalCount);
Collection<IdempotentStorageManager> newManagers =
Lists.newArrayListWithExpectedSize(totalCount);
for (int i = 0; i < scanners.size(); i++) {
// Kryo.copy fails as it attempts to clone transient fields
ByteArrayOutputStream bos = new ByteArrayOutputStream();
Output loutput = new Output(bos);
kryo.writeObject(loutput, this);
loutput.close();
Input lInput = new Input(bos.toByteArray());
@SuppressWarnings("unchecked")
AbstractFileInputOperator<T> oper = kryo.readObject(lInput, this.getClass());
lInput.close();
DirectoryScanner scn = scanners.get(i);
oper.setScanner(scn);
// Do state transfer for processed files.
oper.processedFiles.addAll(totalProcessedFiles);
oper.globalNumberOfFailures = tempGlobalNumberOfRetries;
oper.localNumberOfFailures.setValue(0);
oper.globalNumberOfRetries = tempGlobalNumberOfFailures;
oper.localNumberOfRetries.setValue(0);
/* redistribute unfinished files properly */
oper.unfinishedFiles.clear();
oper.currentFile = null;
oper.offset = 0;
Iterator<FailedFile> unfinishedIter = currentFiles.iterator();
while (unfinishedIter.hasNext()) {
FailedFile unfinishedFile = unfinishedIter.next();
if (scn.acceptFile(unfinishedFile.path)) {
oper.unfinishedFiles.add(unfinishedFile);
unfinishedIter.remove();
}
}
/* transfer failed files */
oper.failedFiles.clear();
Iterator<FailedFile> iter = totalFailedFiles.iterator();
while (iter.hasNext()) {
FailedFile ff = iter.next();
if (scn.acceptFile(ff.path)) {
oper.failedFiles.add(ff);
iter.remove();
}
}
/* redistribute pending files properly */
oper.pendingFiles.clear();
Iterator<String> pendingFilesIterator = totalPendingFiles.iterator();
while (pendingFilesIterator.hasNext()) {
String pathString = pendingFilesIterator.next();
if (scn.acceptFile(pathString)) {
oper.pendingFiles.add(pathString);
pendingFilesIterator.remove();
}
}
newPartitions.add(new DefaultPartition<AbstractFileInputOperator<T>>(oper));
newManagers.add(oper.idempotentStorageManager);
}
idempotentStorageManager.partitioned(newManagers, deletedOperators);
LOG.info("definePartitions called returning {} partitions", newPartitions.size());
return newPartitions;
}
protected void replay(long windowId) {
// This operator can partition itself dynamically. When that happens a file can be re-hashed
// to a different partition than the previous one. In order to handle this, the partition loads
// all the recovery data for a window and then processes only those files which would be hashed
// to it in the current run.
try {
Map<Integer, Object> recoveryDataPerOperator = idempotentStorageManager.load(windowId);
for (Object recovery : recoveryDataPerOperator.values()) {
@SuppressWarnings("unchecked")
LinkedList<RecoveryEntry> recoveryData = (LinkedList<RecoveryEntry>) recovery;
for (RecoveryEntry recoveryEntry : recoveryData) {
if (scanner.acceptFile(recoveryEntry.file)) {
// The operator may have continued processing the same file in multiple windows.
// So the recovery states of subsequent windows will have an entry for that file however
// the offset changes.
// In this case we continue reading from previously opened stream.
if (currentFile == null
|| !(currentFile.equals(recoveryEntry.file)
&& offset == recoveryEntry.startOffset)) {
if (inputStream != null) {
closeFile(inputStream);
}
processedFiles.add(recoveryEntry.file);
// removing the file from failed and unfinished queues and pending set
Iterator<FailedFile> failedFileIterator = failedFiles.iterator();
while (failedFileIterator.hasNext()) {
FailedFile ff = failedFileIterator.next();
if (ff.path.equals(recoveryEntry.file) && ff.offset == recoveryEntry.startOffset) {
failedFileIterator.remove();
break;
}
}
Iterator<FailedFile> unfinishedFileIterator = unfinishedFiles.iterator();
while (unfinishedFileIterator.hasNext()) {
FailedFile ff = unfinishedFileIterator.next();
if (ff.path.equals(recoveryEntry.file) && ff.offset == recoveryEntry.startOffset) {
unfinishedFileIterator.remove();
break;
}
}
if (pendingFiles.contains(recoveryEntry.file)) {
pendingFiles.remove(recoveryEntry.file);
}
inputStream =
retryFailedFile(new FailedFile(recoveryEntry.file, recoveryEntry.startOffset));
while (offset < recoveryEntry.endOffset) {
T line = readEntity();
offset++;
emit(line);
}
} else {
while (offset < recoveryEntry.endOffset) {
T line = readEntity();
offset++;
emit(line);
}
}
}
}
}
} catch (IOException e) {
throw new RuntimeException("replay", e);
}
}