From fe185e2c3a07d4e6a82ed4fd255c98ec8b561ad6 Mon Sep 17 00:00:00 2001 From: Yongjun Zhang Date: Fri, 19 May 2017 21:11:38 -0700 Subject: [PATCH] HADOOP-14407. DistCp - Introduce a configurable copy buffer size. (Omkar Aradhya K S via Yongjun Zhang) --- .../apache/hadoop/tools/DistCpConstants.java | 6 +++ .../hadoop/tools/DistCpOptionSwitch.java | 8 ++++ .../apache/hadoop/tools/DistCpOptions.java | 21 +++++++++- .../apache/hadoop/tools/OptionsParser.java | 27 +++++++++++- .../mapred/RetriableFileCopyCommand.java | 11 +++-- .../src/site/markdown/DistCp.md.vm | 1 + .../hadoop/tools/TestOptionsParser.java | 42 ++++++++++++++++++- 7 files changed, 108 insertions(+), 8 deletions(-) diff --git a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpConstants.java b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpConstants.java index ecf1859ae78..08138468979 100644 --- a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpConstants.java +++ b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpConstants.java @@ -113,6 +113,10 @@ public class DistCpConstants { /* DistCp CopyListing class override param */ public static final String CONF_LABEL_COPY_LISTING_CLASS = "distcp.copy.listing.class"; + /* DistCp Copy Buffer Size */ + public static final String CONF_LABEL_COPY_BUFFER_SIZE = + "distcp.copy.buffer.size"; + /** * Conf label for SSL Trust-store location. */ @@ -157,4 +161,6 @@ public class DistCpConstants { public static final String HDFS_RESERVED_RAW_DIRECTORY_NAME = "/.reserved/raw"; static final String HDFS_DISTCP_DIFF_DIRECTORY_NAME = ".distcp.diff.tmp"; + + public static final int COPY_BUFFER_SIZE_DEFAULT = 8 * 1024; } diff --git a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptionSwitch.java b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptionSwitch.java index e76a48e805f..49ec0351671 100644 --- a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptionSwitch.java +++ b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptionSwitch.java @@ -188,6 +188,14 @@ public enum DistCpOptionSwitch { + "system implements getBlockLocations method and the target file " + "system implements concat method")), + /** + * Configurable copy buffer size. + */ + COPY_BUFFER_SIZE(DistCpConstants.CONF_LABEL_COPY_BUFFER_SIZE, + new Option("copybuffersize", true, "Size of the copy buffer to use. " + + "By default is " + + DistCpConstants.COPY_BUFFER_SIZE_DEFAULT + "B.")), + /** * Specify bandwidth per map in MB */ diff --git a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptions.java b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptions.java index 2efb96b2d95..b3c843fd2b5 100644 --- a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptions.java +++ b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/DistCpOptions.java @@ -104,6 +104,11 @@ public class DistCpOptions { // to copy in parallel. Default is 0 and file are not splitted. private int blocksPerChunk = 0; + /** + * The copyBufferSize to use in RetriableFileCopyCommand + */ + private int copyBufferSize = DistCpConstants.COPY_BUFFER_SIZE_DEFAULT; + public static enum FileAttribute{ REPLICATION, BLOCKSIZE, USER, GROUP, PERMISSION, CHECKSUMTYPE, ACL, XATTR, TIMES; @@ -174,6 +179,7 @@ public DistCpOptions(DistCpOptions that) { this.targetPathExists = that.getTargetPathExists(); this.filtersFile = that.getFiltersFile(); this.blocksPerChunk = that.blocksPerChunk; + this.copyBufferSize = that.copyBufferSize; } } @@ -464,7 +470,7 @@ public Iterator preserveAttributes() { } /** - * Checks if the input attribute should be preserved or not + * Checks if the input attribute should be preserved or not. * * @param attribute - Attribute to check * @return True if attribute should be preserved, false otherwise @@ -640,6 +646,16 @@ public final boolean splitLargeFile() { return blocksPerChunk > 0; } + public final void setCopyBufferSize(int newCopyBufferSize) { + this.copyBufferSize = + newCopyBufferSize > 0 ? newCopyBufferSize + : DistCpConstants.COPY_BUFFER_SIZE_DEFAULT; + } + + public int getCopyBufferSize() { + return this.copyBufferSize; + } + public void validate(DistCpOptionSwitch option, boolean value) { boolean syncFolder = (option == DistCpOptionSwitch.SYNC_FOLDERS ? @@ -736,6 +752,8 @@ public void appendToConf(Configuration conf) { } DistCpOptionSwitch.addToConf(conf, DistCpOptionSwitch.BLOCKS_PER_CHUNK, String.valueOf(blocksPerChunk)); + DistCpOptionSwitch.addToConf(conf, DistCpOptionSwitch.COPY_BUFFER_SIZE, + String.valueOf(copyBufferSize)); } /** @@ -773,6 +791,7 @@ public String toString() { ", targetPathExists=" + targetPathExists + ", filtersFile='" + filtersFile + '\'' + ", blocksPerChunk=" + blocksPerChunk + + ", copyBufferSize=" + copyBufferSize + '}'; } diff --git a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/OptionsParser.java b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/OptionsParser.java index c68102df329..868ae7359b3 100644 --- a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/OptionsParser.java +++ b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/OptionsParser.java @@ -186,6 +186,8 @@ public static DistCpOptions parse(String[] args) parseBlocksPerChunk(command, option); + parseCopyBufferSize(command, option); + return option; } @@ -221,8 +223,29 @@ private static void parseBlocksPerChunk(CommandLine command, } /** - * parseSizeLimit is a helper method for parsing the deprecated - * argument SIZE_LIMIT. + * A helper method to parse copyBufferSize. + * + * @param command command line arguments + */ + private static void parseCopyBufferSize(CommandLine command, + DistCpOptions option) { + if (command.hasOption(DistCpOptionSwitch.COPY_BUFFER_SIZE.getSwitch())) { + String copyBufferSizeStr = + getVal(command, DistCpOptionSwitch.COPY_BUFFER_SIZE.getSwitch() + .trim()); + try { + int copyBufferSize = Integer.parseInt(copyBufferSizeStr); + option.setCopyBufferSize(copyBufferSize); + } catch (NumberFormatException e) { + throw new IllegalArgumentException("copyBufferSize is invalid: " + + copyBufferSizeStr, e); + } + } + } + + /** + * parseSizeLimit is a helper method for parsing the deprecated argument + * SIZE_LIMIT. * * @param command command line arguments */ diff --git a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/mapred/RetriableFileCopyCommand.java b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/mapred/RetriableFileCopyCommand.java index 58a51af15c5..ddf27250e0e 100644 --- a/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/mapred/RetriableFileCopyCommand.java +++ b/hadoop-tools/hadoop-distcp/src/main/java/org/apache/hadoop/tools/mapred/RetriableFileCopyCommand.java @@ -38,6 +38,7 @@ import org.apache.hadoop.mapreduce.Mapper; import org.apache.hadoop.tools.CopyListingFileStatus; import org.apache.hadoop.tools.DistCpConstants; +import org.apache.hadoop.tools.DistCpOptionSwitch; import org.apache.hadoop.tools.DistCpOptions.FileAttribute; import org.apache.hadoop.tools.mapred.CopyMapper.FileAction; import org.apache.hadoop.tools.util.DistCpUtils; @@ -53,7 +54,6 @@ public class RetriableFileCopyCommand extends RetriableCommand { private static Log LOG = LogFactory.getLog(RetriableFileCopyCommand.class); - private static int BUFFER_SIZE = 8 * 1024; private boolean skipCrc = false; private FileAction action; @@ -169,6 +169,9 @@ private long copyToFile(Path targetPath, FileSystem targetFS, throws IOException { FsPermission permission = FsPermission.getFileDefault().applyUMask( FsPermission.getUMask(targetFS.getConf())); + int copyBufferSize = context.getConfiguration().getInt( + DistCpOptionSwitch.COPY_BUFFER_SIZE.getConfigLabel(), + DistCpConstants.COPY_BUFFER_SIZE_DEFAULT); final OutputStream outStream; if (action == FileAction.OVERWRITE) { final short repl = getReplicationFactor(fileAttributes, source, @@ -177,14 +180,14 @@ private long copyToFile(Path targetPath, FileSystem targetFS, targetFS, targetPath); FSDataOutputStream out = targetFS.create(targetPath, permission, EnumSet.of(CreateFlag.CREATE, CreateFlag.OVERWRITE), - BUFFER_SIZE, repl, blockSize, context, + copyBufferSize, repl, blockSize, context, getChecksumOpt(fileAttributes, sourceChecksum)); outStream = new BufferedOutputStream(out); } else { outStream = new BufferedOutputStream(targetFS.append(targetPath, - BUFFER_SIZE)); + copyBufferSize)); } - return copyBytes(source, sourceOffset, outStream, BUFFER_SIZE, + return copyBytes(source, sourceOffset, outStream, copyBufferSize, context); } diff --git a/hadoop-tools/hadoop-distcp/src/site/markdown/DistCp.md.vm b/hadoop-tools/hadoop-distcp/src/site/markdown/DistCp.md.vm index e6cff10b39d..32c71dd16bc 100644 --- a/hadoop-tools/hadoop-distcp/src/site/markdown/DistCp.md.vm +++ b/hadoop-tools/hadoop-distcp/src/site/markdown/DistCp.md.vm @@ -240,6 +240,7 @@ Flag | Description | Notes `-numListstatusThreads` | Number of threads to use for building file listing | At most 40 threads. `-skipcrccheck` | Whether to skip CRC checks between source and target paths. | `-blocksperchunk ` | Number of blocks per chunk. When specified, split files into chunks to copy in parallel | If set to a positive value, files with more blocks than this value will be split into chunks of `` blocks to be transferred in parallel, and reassembled on the destination. By default, `` is 0 and the files will be transmitted in their entirety without splitting. This switch is only applicable when the source file system implements getBlockLocations method and the target file system implements concat method. | +`-copybuffersize ` | Size of the copy buffer to use. By default, `` is set to 8192B | Architecture of DistCp ---------------------- diff --git a/hadoop-tools/hadoop-distcp/src/test/java/org/apache/hadoop/tools/TestOptionsParser.java b/hadoop-tools/hadoop-distcp/src/test/java/org/apache/hadoop/tools/TestOptionsParser.java index acffb76a46e..d1ef56a2919 100644 --- a/hadoop-tools/hadoop-distcp/src/test/java/org/apache/hadoop/tools/TestOptionsParser.java +++ b/hadoop-tools/hadoop-distcp/src/test/java/org/apache/hadoop/tools/TestOptionsParser.java @@ -407,7 +407,8 @@ public void testToString() { + "copyStrategy='uniformsize', preserveStatus=[], " + "preserveRawXattrs=false, atomicWorkPath=null, logPath=null, " + "sourceFileListing=abc, sourcePaths=null, targetPath=xyz, " - + "targetPathExists=true, filtersFile='null', blocksPerChunk=0}"; + + "targetPathExists=true, filtersFile='null', blocksPerChunk=0, " + + "copyBufferSize=8192}"; String optionString = option.toString(); Assert.assertEquals(val, optionString); Assert.assertNotSame(DistCpOptionSwitch.ATOMIC_COMMIT.toString(), @@ -773,4 +774,43 @@ public void testExclusionsOption() { "hdfs://localhost:8020/target/"}); Assert.assertEquals(options.getFiltersFile(), "/tmp/filters.txt"); } + + @Test + public void testParseCopyBufferSize() { + DistCpOptions options = + OptionsParser.parse(new String[] { + "hdfs://localhost:8020/source/first", + "hdfs://localhost:8020/target/" }); + Assert.assertEquals(options.getCopyBufferSize(), + DistCpConstants.COPY_BUFFER_SIZE_DEFAULT); + + options = + OptionsParser.parse(new String[] { "-copybuffersize", "0", + "hdfs://localhost:8020/source/first", + "hdfs://localhost:8020/target/" }); + Assert.assertEquals(options.getCopyBufferSize(), + DistCpConstants.COPY_BUFFER_SIZE_DEFAULT); + + options = + OptionsParser.parse(new String[] { "-copybuffersize", "-1", + "hdfs://localhost:8020/source/first", + "hdfs://localhost:8020/target/" }); + Assert.assertEquals(options.getCopyBufferSize(), + DistCpConstants.COPY_BUFFER_SIZE_DEFAULT); + + options = + OptionsParser.parse(new String[] { "-copybuffersize", "4194304", + "hdfs://localhost:8020/source/first", + "hdfs://localhost:8020/target/" }); + Assert.assertEquals(options.getCopyBufferSize(), 4194304); + + try { + OptionsParser + .parse(new String[] { "-copybuffersize", "hello", + "hdfs://localhost:8020/source/first", + "hdfs://localhost:8020/target/" }); + Assert.fail("Non numberic copybuffersize parsed successfully!"); + } catch (IllegalArgumentException ignore) { + } + } }