Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HDFS-14617. Improve oiv tool to parse fsimage file in parallel with delimited format. (#2918). Contributed by Hongbing Wang. 

Signed-off-by: He Xiaoqiao <hexiaoqiao@apache.org>
This commit is contained in:
He Xiaoqiao 2022-03-22 22:28:09 +08:00
parent 708a0ce21b
commit 88975496d8
No known key found for this signature in database
GPG Key ID: A80CC124E9A0FA63
6 changed files with 306 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/OfflineImageViewerPB.java
View File

@ -107,6 +107,7 @@ public class OfflineImageViewerPB {
+ " Delimited outputs. If not set, the processor\n"
+ " constructs the namespace in memory \n"
+ " before outputting text.\n"
+ "-m,--multiThread <arg> Use multiThread to process sub-sections.\n"
+ "-h,--help Display usage information and exit\n";
/**
@ -132,6 +133,7 @@ public class OfflineImageViewerPB {
options.addOption("delimiter", true, "");
options.addOption("sp", false, "");
options.addOption("t", "temp", true, "");
options.addOption("m", "multiThread", true, "");
return options;
}
@ -185,6 +187,7 @@ public class OfflineImageViewerPB {
String delimiter = cmd.getOptionValue("delimiter",
PBImageTextWriter.DEFAULT_DELIMITER);
String tempPath = cmd.getOptionValue("t", "");
int threads = Integer.parseInt(cmd.getOptionValue("m", "1"));
Configuration conf = new Configuration();
PrintStream out = null;
@ -227,15 +230,14 @@ public class OfflineImageViewerPB {
boolean printStoragePolicy = cmd.hasOption("sp");
try (PBImageDelimitedTextWriter writer =
new PBImageDelimitedTextWriter(out, delimiter,
tempPath, printStoragePolicy);
RandomAccessFile r = new RandomAccessFile(inputFile, "r")) {
writer.visit(r);
tempPath, printStoragePolicy, threads, outputFile)) {
writer.visit(inputFile);
}
break;
case "DETECTCORRUPTION":
try (PBImageCorruptionDetector detector =
new PBImageCorruptionDetector(out, delimiter, tempPath)) {
detector.visit(new RandomAccessFile(inputFile, "r"));
detector.visit(inputFile);
}
break;
default:

2
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/PBImageCorruptionDetector.java
View File

@ -337,7 +337,7 @@ public class PBImageCorruptionDetector extends PBImageTextWriter {
if (parentId != -1) {
entryBuilder.setParentId(parentId);
}
printIfNotEmpty(entryBuilder.build());
printIfNotEmpty(serialOutStream(), entryBuilder.build());
}
}
}

8
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/PBImageDelimitedTextWriter.java
View File

@ -146,7 +146,13 @@ public class PBImageDelimitedTextWriter extends PBImageTextWriter {
PBImageDelimitedTextWriter(PrintStream out, String delimiter,
String tempPath, boolean printStoragePolicy)
throws IOException {
super(out, delimiter, tempPath);
this(out, delimiter, tempPath, printStoragePolicy, 1, "-");
}
PBImageDelimitedTextWriter(PrintStream out, String delimiter,
String tempPath, boolean printStoragePolicy, int threads,
String parallelOut) throws IOException {
super(out, delimiter, tempPath, threads, parallelOut);
this.printStoragePolicy = printStoragePolicy;
}

243
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/PBImageTextWriter.java
View File

@ -21,17 +21,25 @@ import java.io.BufferedInputStream;
import java.io.Closeable;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.PrintStream;
import java.io.RandomAccessFile;
import java.io.UnsupportedEncodingException;
import java.nio.ByteBuffer;
import java.nio.channels.FileChannel;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
import java.util.concurrent.CopyOnWriteArrayList;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
import java.util.HashMap;
import java.util.Iterator;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
@ -455,8 +463,9 @@ abstract class PBImageTextWriter implements Closeable {
return "/";
}
long parent = getFromDirChildMap(inode);
if (!dirPathCache.containsKey(parent)) {
byte[] bytes = dirMap.get(toBytes(parent));
synchronized (this) {
if (!dirPathCache.containsKey(parent)) {
if (parent != INodeId.ROOT_INODE_ID && bytes == null) {
// The parent is an INodeReference, which is generated from snapshot.
// For delimited oiv tool, no need to print out metadata in snapshots.
@ -470,6 +479,7 @@ abstract class PBImageTextWriter implements Closeable {
}
return dirPathCache.get(parent);
}
}
@Override
public void sync() throws IOException {
@ -493,9 +503,12 @@ abstract class PBImageTextWriter implements Closeable {
}
private SerialNumberManager.StringTable stringTable;
private PrintStream out;
private final PrintStream out;
private MetadataMap metadataMap = null;
private String delimiter;
private File filename;
private int numThreads;
private String parallelOutputFile;
/**
* Construct a PB FsImage writer to generate text file.
@ -503,8 +516,8 @@ abstract class PBImageTextWriter implements Closeable {
* @param tempPath the path to store metadata. If it is empty, store metadata
* in memory instead.
*/
PBImageTextWriter(PrintStream out, String delimiter, String tempPath)
throws IOException {
PBImageTextWriter(PrintStream out, String delimiter, String tempPath,
int numThreads, String parallelOutputFile) throws IOException {
this.out = out;
this.delimiter = delimiter;
if (tempPath.isEmpty()) {
@ -512,6 +525,17 @@ abstract class PBImageTextWriter implements Closeable {
} else {
metadataMap = new LevelDBMetadataMap(tempPath);
}
this.numThreads = numThreads;
this.parallelOutputFile = parallelOutputFile;
}
PBImageTextWriter(PrintStream out, String delimiter, String tempPath)
throws IOException {
this(out, delimiter, tempPath, 1, "-");
}
protected PrintStream serialOutStream() {
return out;
}
@Override
@ -562,7 +586,9 @@ abstract class PBImageTextWriter implements Closeable {
*/
abstract protected void afterOutput() throws IOException;
public void visit(RandomAccessFile file) throws IOException {
public void visit(String filePath) throws IOException {
filename = new File(filePath);
RandomAccessFile file = new RandomAccessFile(filePath, "r");
Configuration conf = new Configuration();
if (!FSImageUtil.checkFileFormat(file)) {
throw new IOException("Unrecognized FSImage");
@ -642,16 +668,34 @@ abstract class PBImageTextWriter implements Closeable {
private void output(Configuration conf, FileSummary summary,
FileInputStream fin, ArrayList<FileSummary.Section> sections)
throws IOException {
ArrayList<FileSummary.Section> allINodeSubSections =
getINodeSubSections(sections);
if (numThreads > 1 && !parallelOutputFile.equals("-") &&
allINodeSubSections.size() > 1) {
outputInParallel(conf, summary, allINodeSubSections);
} else {
LOG.info("Serial output due to threads num: {}, parallel output file: {}, " +
"subSections: {}.", numThreads, parallelOutputFile, allINodeSubSections.size());
outputInSerial(conf, summary, fin, sections);
}
}
private void outputInSerial(Configuration conf, FileSummary summary,
FileInputStream fin, ArrayList<FileSummary.Section> sections)
throws IOException {
InputStream is;
long startTime = Time.monotonicNow();
out.println(getHeader());
serialOutStream().println(getHeader());
for (FileSummary.Section section : sections) {
if (SectionName.fromString(section.getName()) == SectionName.INODE) {
fin.getChannel().position(section.getOffset());
is = FSImageUtil.wrapInputStreamForCompression(conf,
summary.getCodec(), new BufferedInputStream(new LimitInputStream(
fin, section.getLength())));
outputINodes(is);
INodeSection s = INodeSection.parseDelimitedFrom(is);
LOG.info("Found {} INodes in the INode section", s.getNumInodes());
int count = outputINodes(is, serialOutStream());
LOG.info("Outputted {} INodes.", count);
}
}
afterOutput();
@ -659,6 +703,89 @@ abstract class PBImageTextWriter implements Closeable {
LOG.debug("Time to output inodes: {} ms", timeTaken);
}
/**
* STEP1: Multi-threaded process sub-sections.
* Given n (n>1) threads to process k (k>=n) sections,
* output parsed results of each section to tmp file in order.
* STEP2: Merge tmp files.
*/
private void outputInParallel(Configuration conf, FileSummary summary,
ArrayList<FileSummary.Section> subSections)
throws IOException {
int nThreads = Integer.min(numThreads, subSections.size());
LOG.info("Outputting in parallel with {} sub-sections using {} threads",
subSections.size(), nThreads);
final CopyOnWriteArrayList<IOException> exceptions = new CopyOnWriteArrayList<>();
CountDownLatch latch = new CountDownLatch(subSections.size());
ExecutorService executorService = Executors.newFixedThreadPool(nThreads);
AtomicLong expectedINodes = new AtomicLong(0);
AtomicLong totalParsed = new AtomicLong(0);
String codec = summary.getCodec();
String[] paths = new String[subSections.size()];
for (int i = 0; i < subSections.size(); i++) {
paths[i] = parallelOutputFile + ".tmp." + i;
int index = i;
executorService.submit(() -> {
LOG.info("Output iNodes of section-{}", index);
InputStream is = null;
try (PrintStream outStream = new PrintStream(paths[index], "UTF-8")) {
long startTime = Time.monotonicNow();
is = getInputStreamForSection(subSections.get(index), codec, conf);
if (index == 0) {
// The first iNode section has a header which must be processed first
INodeSection s = INodeSection.parseDelimitedFrom(is);
expectedINodes.set(s.getNumInodes());
}
totalParsed.addAndGet(outputINodes(is, outStream));
long timeTaken = Time.monotonicNow() - startTime;
LOG.info("Time to output iNodes of section-{}: {} ms", index, timeTaken);
} catch (Exception e) {
exceptions.add(new IOException(e));
} finally {
latch.countDown();
try {
if (is != null) {
is.close();
}
} catch (IOException ioe) {
LOG.warn("Failed to close the input stream, ignoring", ioe);
}
}
});
}
try {
latch.await();
} catch (InterruptedException e) {
LOG.error("Interrupted waiting for countdown latch", e);
throw new IOException(e);
}
executorService.shutdown();
if (exceptions.size() != 0) {
LOG.error("Failed to output INode sub-sections, {} exception(s) occurred.",
exceptions.size());
throw exceptions.get(0);
}
if (totalParsed.get() != expectedINodes.get()) {
throw new IOException("Expected to parse " + expectedINodes + " in parallel, " +
"but parsed " + totalParsed.get() + ". The image may be corrupt.");
}
LOG.info("Completed outputting all INode sub-sections to {} tmp files.",
subSections.size());
try (PrintStream ps = new PrintStream(parallelOutputFile, "UTF-8")) {
ps.println(getHeader());
}
// merge tmp files
long startTime = Time.monotonicNow();
mergeFiles(paths, parallelOutputFile);
long timeTaken = Time.monotonicNow() - startTime;
LOG.info("Completed all stages. Time to merge files: {} ms", timeTaken);
}
protected PermissionStatus getPermission(long perm) {
return FSImageFormatPBINode.Loader.loadPermission(perm, stringTable);
}
@ -763,22 +890,27 @@ abstract class PBImageTextWriter implements Closeable {
LOG.info("Scanned {} INode directories to build namespace.", count);
}
void printIfNotEmpty(String line) {
void printIfNotEmpty(PrintStream outStream, String line) {
if (!line.isEmpty()) {
out.println(line);
outStream.println(line);
}
}
private void outputINodes(InputStream in) throws IOException {
INodeSection s = INodeSection.parseDelimitedFrom(in);
LOG.info("Found {} INodes in the INode section", s.getNumInodes());
private int outputINodes(InputStream in, PrintStream outStream)
throws IOException {
long ignored = 0;
long ignoredSnapshots = 0;
for (int i = 0; i < s.getNumInodes(); ++i) {
// As the input stream is a LimitInputStream, the reading will stop when
// EOF is encountered at the end of the stream.
int count = 0;
while (true) {
INode p = INode.parseDelimitedFrom(in);
if (p == null) {
break;
}
try {
String parentPath = metadataMap.getParentPath(p.getId());
printIfNotEmpty(getEntry(parentPath, p));
printIfNotEmpty(outStream, getEntry(parentPath, p));
} catch (IOException ioe) {
ignored++;
if (!(ioe instanceof IgnoreSnapshotException)) {
@ -790,16 +922,16 @@ abstract class PBImageTextWriter implements Closeable {
}
}
}
if (LOG.isDebugEnabled() && i % 100000 == 0) {
LOG.debug("Outputted {} INodes.", i);
count++;
if (LOG.isDebugEnabled() && count % 100000 == 0) {
LOG.debug("Outputted {} INodes.", count);
}
}
if (ignored > 0) {
LOG.warn("Ignored {} nodes, including {} in snapshots. Please turn on"
+ " debug log for details", ignored, ignoredSnapshots);
}
LOG.info("Outputted {} INodes.", s.getNumInodes());
return count;
}
private static IgnoreSnapshotException createIgnoredSnapshotException(
@ -822,4 +954,79 @@ abstract class PBImageTextWriter implements Closeable {
}
return HdfsConstants.BLOCK_STORAGE_POLICY_ID_UNSPECIFIED;
}
private ArrayList<FileSummary.Section> getINodeSubSections(
ArrayList<FileSummary.Section> sections) {
ArrayList<FileSummary.Section> subSections = new ArrayList<>();
Iterator<FileSummary.Section> iter = sections.iterator();
while (iter.hasNext()) {
FileSummary.Section s = iter.next();
if (SectionName.fromString(s.getName()) == SectionName.INODE_SUB) {
subSections.add(s);
}
}
return subSections;
}
/**
* Given a FSImage FileSummary.section, return a LimitInput stream set to
* the starting position of the section and limited to the section length.
* @param section The FileSummary.Section containing the offset and length
* @param compressionCodec The compression codec in use, if any
* @return An InputStream for the given section
* @throws IOException
*/
private InputStream getInputStreamForSection(FileSummary.Section section,
String compressionCodec, Configuration conf)
throws IOException {
// channel of RandomAccessFile is not thread safe, use File
FileInputStream fin = new FileInputStream(filename);
try {
FileChannel channel = fin.getChannel();
channel.position(section.getOffset());
InputStream in = new BufferedInputStream(new LimitInputStream(fin,
section.getLength()));
in = FSImageUtil.wrapInputStreamForCompression(conf,
compressionCodec, in);
return in;
} catch (IOException e) {
fin.close();
throw e;
}
}
/**
* @param srcPaths Source files of contents to be merged
* @param resultPath Merged file path
* @throws IOException
*/
public static void mergeFiles(String[] srcPaths, String resultPath)
throws IOException {
if (srcPaths == null || srcPaths.length < 1) {
LOG.warn("no source files to merge.");
return;
}
File[] files = new File[srcPaths.length];
for (int i = 0; i < srcPaths.length; i++) {
files[i] = new File(srcPaths[i]);
}
File resultFile = new File(resultPath);
try (FileChannel resultChannel =
new FileOutputStream(resultFile, true).getChannel()) {
for (File file : files) {
try (FileChannel src = new FileInputStream(file).getChannel()) {
resultChannel.transferFrom(src, resultChannel.size(), src.size());
}
}
}
for (File file : files) {
if (!file.delete() && file.exists()) {
LOG.warn("delete tmp file: {} returned false", file);
}
}
}
}

59
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/TestOfflineImageViewer.java
View File

@ -83,8 +83,10 @@ import org.apache.hadoop.hdfs.server.namenode.FSImageTestUtil;
import org.apache.hadoop.hdfs.server.namenode.FsImageProto;
import org.apache.hadoop.hdfs.server.namenode.INodeFile;
import org.apache.hadoop.hdfs.server.namenode.NameNodeLayoutVersion;
import org.apache.hadoop.hdfs.util.MD5FileUtils;
import org.apache.hadoop.hdfs.web.WebHdfsFileSystem;
import org.apache.hadoop.io.IOUtils;
import org.apache.hadoop.io.MD5Hash;
import org.apache.hadoop.io.erasurecode.ECSchema;
import org.apache.hadoop.io.erasurecode.ErasureCodeConstants;
import org.apache.hadoop.net.NetUtils;
@ -122,6 +124,7 @@ import static org.apache.hadoop.fs.permission.AclEntryType.USER;
import static org.apache.hadoop.fs.permission.FsAction.ALL;
import static org.apache.hadoop.fs.permission.FsAction.EXECUTE;
import static org.apache.hadoop.fs.permission.FsAction.READ_EXECUTE;
import static org.apache.hadoop.hdfs.MiniDFSCluster.HDFS_MINIDFS_BASEDIR;
import static org.apache.hadoop.hdfs.server.namenode.AclTestHelpers.aclEntry;
import static org.apache.hadoop.hdfs.tools.offlineImageViewer.PBImageXmlWriter.ERASURE_CODING_SECTION_NAME;
import static org.apache.hadoop.hdfs.tools.offlineImageViewer.PBImageXmlWriter.ERASURE_CODING_SECTION_POLICY;
@ -186,6 +189,12 @@ public class TestOfflineImageViewer {
conf.setBoolean(DFSConfigKeys.DFS_NAMENODE_ACLS_ENABLED_KEY, true);
conf.set(CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTH_TO_LOCAL,
"RULE:[2:$1@$0](JobTracker@.*FOO.COM)s/@.*//" + "DEFAULT");
// fsimage with sub-section conf
conf.set(DFSConfigKeys.DFS_IMAGE_PARALLEL_LOAD_KEY, "true");
conf.set(DFSConfigKeys.DFS_IMAGE_PARALLEL_INODE_THRESHOLD_KEY, "1");
conf.set(DFSConfigKeys.DFS_IMAGE_PARALLEL_TARGET_SECTIONS_KEY, "4");
conf.set(DFSConfigKeys.DFS_IMAGE_PARALLEL_THREADS_KEY, "4");
cluster = new MiniDFSCluster.Builder(conf).numDataNodes(3).build();
cluster.waitActive();
DistributedFileSystem hdfs = cluster.getFileSystem();
@ -791,6 +800,13 @@ public class TestOfflineImageViewer {
new FileSystemTestHelper().getTestRootDir() + "/delimited.db");
}
@Test
public void testParallelPBDelimitedWriter() throws Exception {
testParallelPBDelimitedWriter(""); // Test in memory db.
testParallelPBDelimitedWriter(new FileSystemTestHelper().getTestRootDir()
+ "/parallel-delimited.db");
}
@Test
public void testCorruptionOutputEntryBuilder() throws IOException {
PBImageCorruptionDetector corrDetector =
@ -882,11 +898,10 @@ public class TestOfflineImageViewer {
final String DELIMITER = "\t";
ByteArrayOutputStream output = new ByteArrayOutputStream();
try (PrintStream o = new PrintStream(output);
RandomAccessFile r = new RandomAccessFile(originalFsimage, "r")) {
try (PrintStream o = new PrintStream(output)) {
PBImageDelimitedTextWriter v =
new PBImageDelimitedTextWriter(o, DELIMITER, db);
v.visit(r);
v.visit(originalFsimage.getAbsolutePath());
}
Set<String> fileNames = new HashSet<>();
@ -920,6 +935,37 @@ public class TestOfflineImageViewer {
assertEquals(writtenFiles.keySet(), fileNames);
}
private void testParallelPBDelimitedWriter(String db) throws Exception{
String delimiter = "\t";
int numThreads = 4;
File parallelDelimitedOut = new File(tempDir, "parallelDelimitedOut");
if (OfflineImageViewerPB.run(new String[] {"-p", "Delimited",
"-i", originalFsimage.getAbsolutePath(),
"-o", parallelDelimitedOut.getAbsolutePath(),
"-delimiter", delimiter,
"-t", db,
"-m", String.valueOf(numThreads)}) != 0) {
throw new IOException("oiv returned failure outputting in parallel.");
}
MD5Hash parallelMd5 = MD5FileUtils.computeMd5ForFile(parallelDelimitedOut);
File serialDelimitedOut = new File(tempDir, "serialDelimitedOut");
if (db != "") {
db = db + "/../serial.db";
}
if (OfflineImageViewerPB.run(new String[] {"-p", "Delimited",
"-i", originalFsimage.getAbsolutePath(),
"-o", serialDelimitedOut.getAbsolutePath(),
"-t", db,
"-delimiter", delimiter}) != 0) {
throw new IOException("oiv returned failure outputting in serial.");
}
MD5Hash serialMd5 = MD5FileUtils.computeMd5ForFile(serialDelimitedOut);
assertEquals(parallelMd5, serialMd5);
}
private void testPBCorruptionDetector(String db)
throws IOException, InterruptedException {
final String delimiter = "\t";
@ -928,7 +974,7 @@ public class TestOfflineImageViewer {
try (PrintStream o = new PrintStream(output)) {
PBImageCorruptionDetector v =
new PBImageCorruptionDetector(o, delimiter, db);
v.visit(new RandomAccessFile(originalFsimage, "r"));
v.visit(originalFsimage.getAbsolutePath());
}
try (
@ -1024,7 +1070,7 @@ public class TestOfflineImageViewer {
try (PrintStream o = new PrintStream(output)) {
PBImageCorruptionDetector v =
new PBImageCorruptionDetector(o, ",", db);
v.visit(new RandomAccessFile(corruptedImage, "r"));
v.visit(corruptedImage.getAbsolutePath());
}
return output.toString();
}
@ -1212,6 +1258,9 @@ public class TestOfflineImageViewer {
public void testFileDistributionCalculatorForException() throws Exception {
File fsimageFile = null;
Configuration conf = new Configuration();
// Avoid using the same cluster dir to cause the global originalFsimage
// file to be cleared.
conf.set(HDFS_MINIDFS_BASEDIR, GenericTestUtils.getRandomizedTempPath());
HashMap<String, FileStatus> files = Maps.newHashMap();
// Create a initial fsimage file

2
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/TestOfflineImageViewerForAcl.java
View File

@ -239,7 +239,7 @@ public class TestOfflineImageViewerForAcl {
try (PrintStream o = new PrintStream(output)) {
PBImageDelimitedTextWriter v =
new PBImageDelimitedTextWriter(o, DELIMITER, ""); // run in memory.
v.visit(new RandomAccessFile(originalFsimage, "r"));
v.visit(originalFsimage.getAbsolutePath());
}
try (