public void map(
Text key,
LongWritable value,
OutputCollector<Text, LongWritable> collector,
Reporter reporter)
throws IOException {
String name = key.toString();
long size = value.get();
long seed = Long.parseLong(name);
random.setSeed(seed);
reporter.setStatus("opening " + name);
DataInputStream in = new DataInputStream(fs.open(new Path(DATA_DIR, name)));
long read = 0;
try {
while (read < size) {
long remains = size - read;
int n = (remains <= buffer.length) ? (int) remains : buffer.length;
in.readFully(buffer, 0, n);
read += n;
if (fastCheck) {
Arrays.fill(check, (byte) random.nextInt(Byte.MAX_VALUE));
} else {
random.nextBytes(check);
}
if (n != buffer.length) {
Arrays.fill(buffer, n, buffer.length, (byte) 0);
Arrays.fill(check, n, check.length, (byte) 0);
}
assertTrue(Arrays.equals(buffer, check));
reporter.setStatus("reading " + name + "@" + read + "/" + size);
}
} finally {
in.close();
}
collector.collect(new Text("bytes"), new LongWritable(read));
reporter.setStatus("read " + name);
}
public void map(Text key, Text val, OutputCollector<Text, Text> output, Reporter reporter)
throws IOException {
long acc = 0L;
long recs = 0;
final int keydiff = keymax - keymin;
final int valdiff = valmax - valmin;
for (long i = 0L; acc < bytesToWrite; ++i) {
int recacc = 0;
recacc += generateSentence(key, keymin + (0 == keydiff ? 0 : r.nextInt(keydiff)));
recacc += generateSentence(val, valmin + (0 == valdiff ? 0 : r.nextInt(valdiff)));
output.collect(key, val);
++recs;
acc += recacc;
reporter.incrCounter(Counters.BYTES_WRITTEN, recacc);
reporter.incrCounter(Counters.RECORDS_WRITTEN, 1);
reporter.setStatus(acc + "/" + (bytesToWrite - acc) + " bytes");
}
reporter.setStatus("Wrote " + recs + " records");
}
public void map(
Text key, LongWritable value, OutputCollector<K, LongWritable> collector, Reporter reporter)
throws IOException {
String name = key.toString();
long size = value.get();
long seed = Long.parseLong(name);
if (size == 0) return;
reporter.setStatus("opening " + name);
FSDataInputStream in = fs.open(new Path(DATA_DIR, name));
try {
for (int i = 0; i < SEEKS_PER_FILE; i++) {
// generate a random position
long position = Math.abs(random.nextLong()) % size;
// seek file to that position
reporter.setStatus("seeking " + name);
in.seek(position);
byte b = in.readByte();
// check that byte matches
byte checkByte = 0;
// advance random state to that position
random.setSeed(seed);
for (int p = 0; p <= position; p += check.length) {
reporter.setStatus("generating data for " + name);
if (fastCheck) {
checkByte = (byte) random.nextInt(Byte.MAX_VALUE);
} else {
random.nextBytes(check);
checkByte = check[(int) (position % check.length)];
}
}
assertEquals(b, checkByte);
}
} finally {
in.close();
}
}
public void map(
Text key,
LongWritable value,
OutputCollector<Text, LongWritable> collector,
Reporter reporter)
throws IOException {
String name = key.toString();
long size = value.get();
long seed = Long.parseLong(name);
random.setSeed(seed);
reporter.setStatus("creating " + name);
// write to temp file initially to permit parallel execution
Path tempFile = new Path(DATA_DIR, name + suffix);
OutputStream out = fs.create(tempFile);
long written = 0;
try {
while (written < size) {
if (fastCheck) {
Arrays.fill(buffer, (byte) random.nextInt(Byte.MAX_VALUE));
} else {
random.nextBytes(buffer);
}
long remains = size - written;
int length = (remains <= buffer.length) ? (int) remains : buffer.length;
out.write(buffer, 0, length);
written += length;
reporter.setStatus("writing " + name + "@" + written + "/" + size);
}
} finally {
out.close();
}
// rename to final location
fs.rename(tempFile, new Path(DATA_DIR, name));
collector.collect(new Text("bytes"), new LongWritable(written));
reporter.setStatus("wrote " + name);
}
private boolean validate(String str, Reporter reporter) {
String[] parts = str.split("\t");
if (parts.length != 6) {
if (parts.length < 6) {
reporter.incrCounter(LineCounters.TOO_FEW_TABS, 1);
} else {
reporter.incrCounter(LineCounters.TOO_MANY_TABS, 1);
}
reporter.incrCounter(LineCounters.BAD_LINES, 1);
if ((reporter.getCounter(LineCounters.BAD_LINES).getCounter() % 10) == 0) {
reporter.setStatus("Got 10 bad lines.");
System.err.println("Read another 10 bad lines.");
}
return false;
}
return true;
}
public Long doIO(Reporter reporter, String name, long totalSize) throws IOException {
totalSize *= MEGA;
// create instance of local filesystem
FileSystem localFS = FileSystem.getLocal(fsConfig);
try {
// native runtime
Runtime runTime = Runtime.getRuntime();
// copy the dso and executable from dfs
synchronized (this) {
localFS.delete(HDFS_TEST_DIR, true);
if (!(localFS.mkdirs(HDFS_TEST_DIR))) {
throw new IOException("Failed to create " + HDFS_TEST_DIR + " on local filesystem");
}
}
synchronized (this) {
if (!localFS.exists(HDFS_SHLIB)) {
if (!FileUtil.copy(fs, HDFS_SHLIB, localFS, HDFS_SHLIB, false, fsConfig)) {
throw new IOException("Failed to copy " + HDFS_SHLIB + " to local filesystem");
}
String chmodCmd = new String(CHMOD + " a+x " + HDFS_SHLIB);
Process process = runTime.exec(chmodCmd);
int exitStatus = process.waitFor();
if (exitStatus != 0) {
throw new IOException(chmodCmd + ": Failed with exitStatus: " + exitStatus);
}
}
}
synchronized (this) {
if (!localFS.exists(HDFS_READ)) {
if (!FileUtil.copy(fs, HDFS_READ, localFS, HDFS_READ, false, fsConfig)) {
throw new IOException("Failed to copy " + HDFS_READ + " to local filesystem");
}
String chmodCmd = new String(CHMOD + " a+x " + HDFS_READ);
Process process = runTime.exec(chmodCmd);
int exitStatus = process.waitFor();
if (exitStatus != 0) {
throw new IOException(chmodCmd + ": Failed with exitStatus: " + exitStatus);
}
}
}
// exec the C program
Path inFile = new Path(DATA_DIR, name);
String readCmd = new String(HDFS_READ + " " + inFile + " " + totalSize + " " + bufferSize);
Process process = runTime.exec(readCmd, null, new File(HDFS_TEST_DIR.toString()));
int exitStatus = process.waitFor();
if (exitStatus != 0) {
throw new IOException(HDFS_READ + ": Failed with exitStatus: " + exitStatus);
}
} catch (InterruptedException interruptedException) {
reporter.setStatus(interruptedException.toString());
} finally {
localFS.close();
}
return new Long(totalSize);
}
/**
* Copy a file to a destination.
*
* @param srcstat src path and metadata
* @param dstpath dst path
* @param reporter
*/
private void copy(
FileStatus srcstat,
Path relativedst,
OutputCollector<WritableComparable<?>, Text> outc,
Reporter reporter)
throws IOException {
Path absdst = new Path(destPath, relativedst);
int totfiles = job.getInt(SRC_COUNT_LABEL, -1);
assert totfiles >= 0 : "Invalid file count " + totfiles;
// if a directory, ensure created even if empty
if (srcstat.isDir()) {
if (destFileSys.exists(absdst)) {
if (!destFileSys.getFileStatus(absdst).isDir()) {
throw new IOException("Failed to mkdirs: " + absdst + " is a file.");
}
} else if (!destFileSys.mkdirs(absdst)) {
throw new IOException("Failed to mkdirs " + absdst);
}
// TODO: when modification times can be set, directories should be
// emitted to reducers so they might be preserved. Also, mkdirs does
// not currently return an error when the directory already exists;
// if this changes, all directory work might as well be done in reduce
return;
}
if (destFileSys.exists(absdst) && !overwrite && !needsUpdate(srcstat, destFileSys, absdst)) {
outc.collect(null, new Text("SKIP: " + srcstat.getPath()));
++skipcount;
reporter.incrCounter(Counter.SKIP, 1);
updateStatus(reporter);
return;
}
Path tmpfile = new Path(job.get(TMP_DIR_LABEL), relativedst);
long cbcopied = 0L;
FSDataInputStream in = null;
FSDataOutputStream out = null;
try {
// open src file
try {
in = srcstat.getPath().getFileSystem(job).open(srcstat.getPath());
} catch (IOException e) {
LOG.error("Failed to open src file " + srcstat.getPath() + ", ignore and return");
in = null;
return;
}
reporter.incrCounter(Counter.BYTESEXPECTED, srcstat.getLen());
// open tmp file
out = create(tmpfile, reporter, srcstat);
// copy file
for (int cbread; (cbread = in.read(buffer)) >= 0; ) {
out.write(buffer, 0, cbread);
cbcopied += cbread;
reporter.setStatus(
String.format("%.2f ", cbcopied * 100.0 / srcstat.getLen())
+ absdst
+ " [ "
+ StringUtils.humanReadableInt(cbcopied)
+ " / "
+ StringUtils.humanReadableInt(srcstat.getLen())
+ " ]");
}
} finally {
checkAndClose(in);
checkAndClose(out);
}
if (cbcopied != srcstat.getLen()) {
if (srcstat.getLen() == 0 && cbcopied > 0) {
LOG.info("most likely see a WAL file corruption: " + srcstat.getPath());
} else {
throw new IOException(
"File size not matched: copied "
+ bytesString(cbcopied)
+ " to tmpfile (="
+ tmpfile
+ ") but expected "
+ bytesString(srcstat.getLen())
+ " from "
+ srcstat.getPath());
}
} else {
if (totfiles == 1) {
// Copying a single file; use dst path provided by user as destination
// rather than destination directory, if a file
Path dstparent = absdst.getParent();
if (!(destFileSys.exists(dstparent) && destFileSys.getFileStatus(dstparent).isDir())) {
absdst = dstparent;
}
}
if (destFileSys.exists(absdst) && destFileSys.getFileStatus(absdst).isDir()) {
throw new IOException(absdst + " is a directory");
}
if (!destFileSys.mkdirs(absdst.getParent())) {
throw new IOException("Failed to craete parent dir: " + absdst.getParent());
}
rename(tmpfile, absdst);
FileStatus dststat = destFileSys.getFileStatus(absdst);
if (dststat.getLen() != srcstat.getLen()) {
destFileSys.delete(absdst, false);
throw new IOException(
"File size not matched: copied "
+ bytesString(dststat.getLen())
+ " to dst (="
+ absdst
+ ") but expected "
+ bytesString(srcstat.getLen())
+ " from "
+ srcstat.getPath());
}
updatePermissions(srcstat, dststat);
}
// report at least once for each file
++copycount;
reporter.incrCounter(Counter.BYTESCOPIED, cbcopied);
reporter.incrCounter(Counter.COPY, 1);
updateStatus(reporter);
}
private void updateStatus(Reporter reporter) {
reporter.setStatus(getCountString());
}