use of org.apache.carbondata.format.BlockIndex in project carbondata by apache.
the class StreamSegment method updateIndexFile.
/**
* update carbonindex file after a stream batch.
*/
public static void updateIndexFile(String segmentDir) throws IOException {
FileFactory.FileType fileType = FileFactory.getFileType(segmentDir);
String filePath = CarbonTablePath.getCarbonStreamIndexFilePath(segmentDir);
String tempFilePath = filePath + CarbonCommonConstants.TEMPWRITEFILEEXTENSION;
CarbonIndexFileWriter writer = new CarbonIndexFileWriter();
try {
writer.openThriftWriter(tempFilePath);
CarbonFile[] files = listDataFiles(segmentDir, fileType);
BlockIndex blockIndex;
for (CarbonFile file : files) {
blockIndex = new BlockIndex();
blockIndex.setFile_name(file.getName());
blockIndex.setFile_size(file.getSize());
// TODO need to collect these information
blockIndex.setNum_rows(-1);
blockIndex.setOffset(-1);
blockIndex.setBlock_index(new BlockletIndex());
writer.writeThrift(blockIndex);
}
writer.close();
CarbonFile tempFile = FileFactory.getCarbonFile(tempFilePath, fileType);
if (!tempFile.renameForce(filePath)) {
throw new IOException("temporary file renaming failed, src=" + tempFilePath + ", dest=" + filePath);
}
} catch (IOException ex) {
try {
writer.close();
} catch (IOException t) {
LOGGER.error(t);
}
throw ex;
}
}
use of org.apache.carbondata.format.BlockIndex in project carbondata by apache.
the class StreamSegment method size.
/**
* calculate the size of the segment by the accumulation of data sizes in index file
*/
public static long size(String segmentDir) throws IOException {
long size = 0;
FileFactory.FileType fileType = FileFactory.getFileType(segmentDir);
if (FileFactory.isFileExist(segmentDir, fileType)) {
String indexPath = CarbonTablePath.getCarbonStreamIndexFilePath(segmentDir);
CarbonFile index = FileFactory.getCarbonFile(indexPath, fileType);
if (index.exists()) {
CarbonIndexFileReader indexReader = new CarbonIndexFileReader();
try {
indexReader.openThriftReader(indexPath);
while (indexReader.hasNext()) {
BlockIndex blockIndex = indexReader.readBlockIndexInfo();
size += blockIndex.getFile_size();
}
} finally {
indexReader.closeThriftReader();
}
}
}
return size;
}
use of org.apache.carbondata.format.BlockIndex in project carbondata by apache.
the class StreamSegment method recoverSegmentIfRequired.
/**
* check the health of stream segment and try to recover segment from job fault
* this method will be invoked in following scenarios.
* 1. at the begin of the streaming (StreamSinkFactory.getStreamSegmentId)
* 2. after job failed (CarbonAppendableStreamSink.writeDataFileJob)
*/
public static void recoverSegmentIfRequired(String segmentDir) throws IOException {
FileFactory.FileType fileType = FileFactory.getFileType(segmentDir);
if (FileFactory.isFileExist(segmentDir, fileType)) {
String indexName = CarbonTablePath.getCarbonStreamIndexFileName();
String indexPath = segmentDir + File.separator + indexName;
CarbonFile index = FileFactory.getCarbonFile(indexPath, fileType);
CarbonFile[] files = listDataFiles(segmentDir, fileType);
// index file exists
if (index.exists()) {
// data file exists
if (files.length > 0) {
CarbonIndexFileReader indexReader = new CarbonIndexFileReader();
try {
// map block index
indexReader.openThriftReader(indexPath);
Map<String, Long> tableSizeMap = new HashMap<>();
while (indexReader.hasNext()) {
BlockIndex blockIndex = indexReader.readBlockIndexInfo();
tableSizeMap.put(blockIndex.getFile_name(), blockIndex.getFile_size());
}
// recover each file
for (CarbonFile file : files) {
Long size = tableSizeMap.get(file.getName());
if (null == size || size == 0) {
file.delete();
} else if (size < file.getSize()) {
FileFactory.truncateFile(file.getCanonicalPath(), fileType, size);
}
}
} finally {
indexReader.closeThriftReader();
}
}
} else {
if (files.length > 0) {
for (CarbonFile file : files) {
file.delete();
}
}
}
}
}
use of org.apache.carbondata.format.BlockIndex in project carbondata by apache.
the class StreamSegment method recoverFileIfRequired.
/**
* check the health of stream data file and try to recover data file from task fault
* this method will be invoked in following scenarios.
* 1. at the begin of writing data file task
*/
public static void recoverFileIfRequired(String segmentDir, String fileName, String indexName) throws IOException {
FileFactory.FileType fileType = FileFactory.getFileType(segmentDir);
String filePath = segmentDir + File.separator + fileName;
CarbonFile file = FileFactory.getCarbonFile(filePath, fileType);
String indexPath = segmentDir + File.separator + indexName;
CarbonFile index = FileFactory.getCarbonFile(indexPath, fileType);
if (file.exists() && index.exists()) {
CarbonIndexFileReader indexReader = new CarbonIndexFileReader();
try {
indexReader.openThriftReader(indexPath);
while (indexReader.hasNext()) {
BlockIndex blockIndex = indexReader.readBlockIndexInfo();
if (blockIndex.getFile_name().equals(fileName)) {
if (blockIndex.getFile_size() == 0) {
file.delete();
} else if (blockIndex.getFile_size() < file.getSize()) {
FileFactory.truncateFile(filePath, fileType, blockIndex.getFile_size());
}
}
}
} finally {
indexReader.closeThriftReader();
}
}
}
use of org.apache.carbondata.format.BlockIndex in project carbondata by apache.
the class CarbonTableInputFormat method getSplitsOfStreaming.
/**
* use file list in .carbonindex file to get the split of streaming.
*/
public List<InputSplit> getSplitsOfStreaming(JobContext job, AbsoluteTableIdentifier identifier, List<Segment> streamSegments) throws IOException {
List<InputSplit> splits = new ArrayList<InputSplit>();
if (streamSegments != null && !streamSegments.isEmpty()) {
numStreamSegments = streamSegments.size();
long minSize = Math.max(getFormatMinSplitSize(), getMinSplitSize(job));
long maxSize = getMaxSplitSize(job);
for (Segment segment : streamSegments) {
String segmentDir = CarbonTablePath.getSegmentPath(identifier.getTablePath(), segment.getSegmentNo());
FileFactory.FileType fileType = FileFactory.getFileType(segmentDir);
if (FileFactory.isFileExist(segmentDir, fileType)) {
String indexName = CarbonTablePath.getCarbonStreamIndexFileName();
String indexPath = segmentDir + File.separator + indexName;
CarbonFile index = FileFactory.getCarbonFile(indexPath, fileType);
// index file exists
if (index.exists()) {
// data file exists
CarbonIndexFileReader indexReader = new CarbonIndexFileReader();
try {
// map block index
indexReader.openThriftReader(indexPath);
while (indexReader.hasNext()) {
BlockIndex blockIndex = indexReader.readBlockIndexInfo();
String filePath = segmentDir + File.separator + blockIndex.getFile_name();
Path path = new Path(filePath);
long length = blockIndex.getFile_size();
if (length != 0) {
BlockLocation[] blkLocations;
FileSystem fs = FileFactory.getFileSystem(path);
FileStatus file = fs.getFileStatus(path);
blkLocations = fs.getFileBlockLocations(path, 0, length);
long blockSize = file.getBlockSize();
long splitSize = computeSplitSize(blockSize, minSize, maxSize);
long bytesRemaining = length;
while (((double) bytesRemaining) / splitSize > 1.1) {
int blkIndex = getBlockIndex(blkLocations, length - bytesRemaining);
splits.add(makeSplit(segment.getSegmentNo(), path, length - bytesRemaining, splitSize, blkLocations[blkIndex].getHosts(), blkLocations[blkIndex].getCachedHosts(), FileFormat.ROW_V1));
bytesRemaining -= splitSize;
}
if (bytesRemaining != 0) {
int blkIndex = getBlockIndex(blkLocations, length - bytesRemaining);
splits.add(makeSplit(segment.getSegmentNo(), path, length - bytesRemaining, bytesRemaining, blkLocations[blkIndex].getHosts(), blkLocations[blkIndex].getCachedHosts(), FileFormat.ROW_V1));
}
} else {
// Create empty hosts array for zero length files
splits.add(makeSplit(segment.getSegmentNo(), path, 0, length, new String[0], FileFormat.ROW_V1));
}
}
} finally {
indexReader.closeThriftReader();
}
}
}
}
}
return splits;
}
Aggregations