public void setConf(Configuration conf) {
this.conf = conf;
scorePower = conf.getFloat("indexer.score.power", 0.5f);
internalScoreFactor = conf.getFloat("db.score.link.internal", 1.0f);
externalScoreFactor = conf.getFloat("db.score.link.external", 1.0f);
countFiltered = conf.getBoolean("db.score.count.filtered", false);
}
public void setConf(Configuration conf) {
super.setConf(conf);
if (conf == null) return;
INC_RATE = conf.getFloat("db.fetch.schedule.adaptive.inc_rate", 0.2f);
DEC_RATE = conf.getFloat("db.fetch.schedule.adaptive.dec_rate", 0.2f);
MIN_INTERVAL = conf.getInt("db.fetch.schedule.adaptive.min_interval", 60);
MAX_INTERVAL =
conf.getInt(
"db.fetch.schedule.adaptive.max_interval",
FetchSchedule.SECONDS_PER_DAY * 365); // 1 year
SYNC_DELTA = conf.getBoolean("db.fetch.schedule.adaptive.sync_delta", true);
SYNC_DELTA_RATE = conf.getFloat("db.fetch.schedule.adaptive.sync_delta_rate", 0.2f);
}
@Override
public void setConf(Configuration conf) {
this.conf = conf;
this.maximumStepSize = conf.getFloat(MAX_STEP_KEY, DEFAULT_MAX_STEP_VALUE);
this.minimumStepSize = conf.getFloat(MIN_STEP_KEY, DEFAULT_MIN_STEP_VALUE);
this.step = this.maximumStepSize;
this.sufficientMemoryLevel =
conf.getFloat(SUFFICIENT_MEMORY_LEVEL_KEY, DEFAULT_SUFFICIENT_MEMORY_LEVEL_VALUE);
this.tunerLookupPeriods = conf.getInt(LOOKUP_PERIODS_KEY, DEFAULT_LOOKUP_PERIODS);
this.blockCachePercentMinRange =
conf.getFloat(
BLOCK_CACHE_SIZE_MIN_RANGE_KEY,
conf.getFloat(HFILE_BLOCK_CACHE_SIZE_KEY, HConstants.HFILE_BLOCK_CACHE_SIZE_DEFAULT));
this.blockCachePercentMaxRange =
conf.getFloat(
BLOCK_CACHE_SIZE_MAX_RANGE_KEY,
conf.getFloat(HFILE_BLOCK_CACHE_SIZE_KEY, HConstants.HFILE_BLOCK_CACHE_SIZE_DEFAULT));
this.globalMemStorePercentMinRange =
conf.getFloat(
MEMSTORE_SIZE_MIN_RANGE_KEY, HeapMemorySizeUtil.getGlobalMemStorePercent(conf, false));
this.globalMemStorePercentMaxRange =
conf.getFloat(
MEMSTORE_SIZE_MAX_RANGE_KEY, HeapMemorySizeUtil.getGlobalMemStorePercent(conf, false));
// Default value of periods to ignore is number of lookup periods
this.numPeriodsToIgnore = conf.getInt(NUM_PERIODS_TO_IGNORE, this.tunerLookupPeriods);
this.rollingStatsForCacheMisses = new RollingStatCalculator(this.tunerLookupPeriods);
this.rollingStatsForFlushes = new RollingStatCalculator(this.tunerLookupPeriods);
this.rollingStatsForEvictions = new RollingStatCalculator(this.tunerLookupPeriods);
}
@Override
public void setup(Context context) throws IOException {
Configuration conf = context.getConfiguration();
Path cMemMatrixPath = new Path(conf.get(RECONSTRUCTIONMATRIX));
Path dMemMatrixPath = new Path(conf.get(MATRIXY2X));
Path zmPath = new Path(conf.get(ZMPATH));
Path meanPath = new Path(conf.get(YMPATH));
int inMemMatrixNumRows = conf.getInt(YCOLS, 0);
int inMemMatrixNumCols = conf.getInt(XCOLS, 0);
ERR_SAMPLE_RATE = conf.getFloat(ERRSAMPLERATE, 1);
Path tmpPath = cMemMatrixPath.getParent();
DistributedRowMatrix distMatrix =
new DistributedRowMatrix(cMemMatrixPath, tmpPath, inMemMatrixNumRows, inMemMatrixNumCols);
distMatrix.setConf(conf);
matrixC = PCACommon.toDenseMatrix(distMatrix);
distMatrix =
new DistributedRowMatrix(dMemMatrixPath, tmpPath, inMemMatrixNumRows, inMemMatrixNumCols);
distMatrix.setConf(conf);
matrixY2X = PCACommon.toDenseMatrix(distMatrix);
try {
zm = PCACommon.toDenseVector(zmPath, conf);
ym = PCACommon.toDenseVector(meanPath, conf);
} catch (IOException e) {
e.printStackTrace();
}
xiCt = new DenseVector(matrixC.numRows());
sumOfErr = new DenseVector(matrixC.numRows());
sumOfyi = new DenseVector(matrixC.numRows());
sumOfyc = new DenseVector(matrixC.numRows());
}
/**
* This functions modifies the splits list in place to retain only a random fraction of the input
* splits. The fraction is expected in "starfish.profiler.sampling.fraction" as a number between 0
* and 1. The default value is 0.1.
*
* @param job The job context
* @param splits The list of input splits to modify
*/
public static void sampleInputSplits(JobContext job, List<InputSplit> splits) {
// Get the sampling fraction
Configuration conf = job.getConfiguration();
double fraction = conf.getFloat(Profiler.PROFILER_SAMPLING_FRACTION, 0.1f);
if (fraction < 0 || fraction > 1)
throw new RuntimeException("ERROR: Invalid sampling fraction: " + fraction);
// Handle corner cases
if (fraction == 0 || splits.size() == 0) {
splits.clear();
return;
}
if (fraction == 1) return;
// Calculate the number of samples
int numSplits = splits.size();
int sampleSize = (int) Math.round(numSplits * fraction);
if (sampleSize == 0) sampleSize = 1;
// Shuffle the splits
Collections.shuffle(splits);
// Retain only a sampleSize number of splits
for (int i = splits.size() - 1; i >= sampleSize; --i) {
splits.remove(i);
}
nf.setMaximumFractionDigits(2);
LOG.info("Executing only " + nf.format(fraction * 100) + "% of the map tasks");
}
InputErrorTracker(Configuration conf) {
// default threshold : 0.01%
errorThreshold = conf.getFloat(BAD_RECORD_THRESHOLD_CONF_KEY, 0.0001f);
minErrors = conf.getLong(BAD_RECORD_MIN_COUNT_CONF_KEY, 2);
numRecords = 0;
numErrors = 0;
}
@Override
public void setup(Context context) throws IOException {
Configuration conf = context.getConfiguration();
missingMass = conf.getFloat("MissingMass", 0.0f);
nodeCnt = conf.getInt("NodeCount", 0);
}
/** Start the JobTracker process, listen on the indicated port */
JobTracker(Configuration conf) throws IOException {
//
// Grab some static constants
//
maxCurrentTasks = conf.getInt("mapred.tasktracker.tasks.maximum", 2);
RETIRE_JOB_INTERVAL = conf.getLong("mapred.jobtracker.retirejob.interval", 24 * 60 * 60 * 1000);
RETIRE_JOB_CHECK_INTERVAL = conf.getLong("mapred.jobtracker.retirejob.check", 60 * 1000);
TASK_ALLOC_EPSILON = conf.getFloat("mapred.jobtracker.taskalloc.loadbalance.epsilon", 0.2f);
PAD_FRACTION = conf.getFloat("mapred.jobtracker.taskalloc.capacitypad", 0.1f);
MIN_SLOTS_FOR_PADDING = 3 * maxCurrentTasks;
// This is a directory of temporary submission files. We delete it
// on startup, and can delete any files that we're done with
this.conf = conf;
JobConf jobConf = new JobConf(conf);
this.systemDir = jobConf.getSystemDir();
this.fs = FileSystem.get(conf);
FileUtil.fullyDelete(fs, systemDir);
fs.mkdirs(systemDir);
// Same with 'localDir' except it's always on the local disk.
jobConf.deleteLocalFiles(SUBDIR);
// Set ports, start RPC servers, etc.
InetSocketAddress addr = getAddress(conf);
this.localMachine = addr.getHostName();
this.port = addr.getPort();
this.interTrackerServer = RPC.getServer(this, addr.getPort(), 10, false, conf);
this.interTrackerServer.start();
Properties p = System.getProperties();
for (Iterator it = p.keySet().iterator(); it.hasNext(); ) {
String key = (String) it.next();
String val = (String) p.getProperty(key);
LOG.info("Property '" + key + "' is " + val);
}
this.infoPort = conf.getInt("mapred.job.tracker.info.port", 50030);
this.infoServer = new JobTrackerInfoServer(this, infoPort);
this.infoServer.start();
this.startTime = System.currentTimeMillis();
new Thread(this.expireTrackers).start();
new Thread(this.retireJobs).start();
new Thread(this.initJobs).start();
}
static NaiveBayesModel readModelFromTempDir(Path base, Configuration conf) {
float alphaI = conf.getFloat(ThetaMapper.ALPHA_I, 1.0f);
// read feature sums and label sums
Vector scoresPerLabel = null;
Vector scoresPerFeature = null;
for (Pair<Text, VectorWritable> record :
new SequenceFileDirIterable<Text, VectorWritable>(
new Path(base, TrainNaiveBayesJob.WEIGHTS),
PathType.LIST,
PathFilters.partFilter(),
conf)) {
String key = record.getFirst().toString();
VectorWritable value = record.getSecond();
if (key.equals(TrainNaiveBayesJob.WEIGHTS_PER_FEATURE)) {
scoresPerFeature = value.get();
} else if (key.equals(TrainNaiveBayesJob.WEIGHTS_PER_LABEL)) {
scoresPerLabel = value.get();
}
}
Preconditions.checkNotNull(scoresPerFeature);
Preconditions.checkNotNull(scoresPerLabel);
Matrix scoresPerLabelAndFeature =
new SparseMatrix(scoresPerLabel.size(), scoresPerFeature.size());
for (Pair<IntWritable, VectorWritable> entry :
new SequenceFileDirIterable<IntWritable, VectorWritable>(
new Path(base, TrainNaiveBayesJob.SUMMED_OBSERVATIONS),
PathType.LIST,
PathFilters.partFilter(),
conf)) {
scoresPerLabelAndFeature.assignRow(entry.getFirst().get(), entry.getSecond().get());
}
Vector perlabelThetaNormalizer = null;
for (Pair<Text, VectorWritable> entry :
new SequenceFileDirIterable<Text, VectorWritable>(
new Path(base, TrainNaiveBayesJob.THETAS),
PathType.LIST,
PathFilters.partFilter(),
conf)) {
if (entry.getFirst().toString().equals(TrainNaiveBayesJob.LABEL_THETA_NORMALIZER)) {
perlabelThetaNormalizer = entry.getSecond().get();
}
}
Preconditions.checkNotNull(perlabelThetaNormalizer);
return new NaiveBayesModel(
scoresPerLabelAndFeature,
scoresPerFeature,
scoresPerLabel,
perlabelThetaNormalizer,
alphaI);
}
/* (non-Javadoc)
* @see org.apache.hadoop.mapreduce.Reducer#setup(org.apache.hadoop.mapreduce.Reducer.Context)
*/
protected void setup(Context context) throws IOException, InterruptedException {
Configuration conf = context.getConfiguration();
fieldDelim = conf.get("field.delim", ",");
linearCorrelation = conf.getBoolean("utp.correlation.linear", true);
correlationScale = conf.getInt("utp.correlation.linear.scale", 1000);
maxRating = conf.getInt("utp.max.rating", 100);
correlationModifier = conf.getFloat("utp.correlation.modifier", (float) 1.0);
userRatingWithContext = conf.getBoolean("utp.user.rating.with.context", false);
}
@Override
public void contextualize(Configuration conf, AppContext context) {
this.context = context;
Map<JobId, Job> allJobs = context.getAllJobs();
for (Map.Entry<JobId, Job> entry : allJobs.entrySet()) {
final Job job = entry.getValue();
mapperStatistics.put(job, new DataStatistics());
reducerStatistics.put(job, new DataStatistics());
slowTaskRelativeTresholds.put(
job, conf.getFloat(MRJobConfig.SPECULATIVE_SLOWTASK_THRESHOLD, 1.0f));
}
}
private synchronized void initBloomFilter(Configuration conf) {
numKeys = conf.getInt("io.mapfile.bloom.size", 1024 * 1024);
// vector size should be <code>-kn / (ln(1 - c^(1/k)))</code> bits for
// single key, where <code> is the number of hash functions,
// <code>n</code> is the number of keys and <code>c</code> is the desired
// max. error rate.
// Our desired error rate is by default 0.005, i.e. 0.5%
float errorRate = conf.getFloat("io.mapfile.bloom.error.rate", 0.005f);
vectorSize =
(int)
Math.ceil(
(double) (-HASH_COUNT * numKeys)
/ Math.log(1.0 - Math.pow(errorRate, 1.0 / HASH_COUNT)));
bloomFilter = new DynamicBloomFilter(vectorSize, HASH_COUNT, Hash.getHashType(conf), numKeys);
}
/* (non-Javadoc)
* @see org.chombo.mr.Transformer.TransformerMapper#setup(org.apache.hadoop.mapreduce.Mapper.Context)
*/
protected void setup(Context context) throws IOException, InterruptedException {
super.setup(context);
Configuration config = context.getConfiguration();
String strategy = config.get("novelty.gen.strategy", "selfInformation");
int maxRating = config.getInt("rating.scale", 100);
RedisCache cache = RedisCache.createRedisCache(config, "ch");
if (strategy.equals("selfInformation")) {
// based on rating distribution
String countMaxValueKeyPrefix = config.get("count.max.value.key.prefix");
int engaementDistrScale = cache.getIntMax(countMaxValueKeyPrefix);
registerTransformers(1, new Transformer.NullTransformer());
registerTransformers(
2, new IndividualNovelty.SelfInformation(engaementDistrScale, maxRating));
} else if (strategy.equals("nonLinearInverse")) {
// based on rating
double param = config.getFloat("quadratic.param", (float) 0.8);
String ratingMaxValueKeyPrefix = config.get("rating.max.value.key.prefix");
int maxRatingInData = cache.getIntMax(ratingMaxValueKeyPrefix);
registerTransformers(
1, new IndividualNovelty.NonLinearInverse(maxRating, param, maxRatingInData));
registerTransformers(2, new Transformer.NullTransformer());
}
}
/**
* Enable profiling for only a fraction of the MR tasks. The fraction is expected in
* "starfish.profiler.sampling.fraction" as a number between 0 and 1. The default value is 0.1.
*
* <p>This function sets the Hadoop parameters mapred.task.profile.maps and
* mapred.task.profile.reducers
*
* @param conf The job configuration
* @return False if something goes wrong
*/
public static boolean sampleTasksToProfile(Configuration conf) {
double fraction = conf.getFloat(Profiler.PROFILER_SAMPLING_FRACTION, 0.1f);
if (fraction == 0d) {
// Nothing to profile
return false;
}
List<SFInputSplit> splits = null;
try {
splits = SFInputSplit.getInputSplits(conf);
} catch (IOException e) {
LOG.error("Unable to create input splits", e);
return false;
} catch (InterruptedException e) {
LOG.error("Unable to create input splits", e);
return false;
} catch (ClassNotFoundException e) {
LOG.error("Unable to create input splits", e);
return false;
}
// Specify which mappers to profile
if (splits != null && splits.size() != 0) {
conf.set("mapred.task.profile.maps", sampleTasksToProfile(splits.size(), fraction));
}
// Specify which reducers to profile
int numReducers = conf.getInt(MR_RED_TASKS, 1);
if (numReducers != 0) {
conf.set("mapred.task.profile.reduces", sampleTasksToProfile(numReducers, fraction));
}
nf.setMaximumFractionDigits(2);
LOG.info("Profiling only " + nf.format(fraction * 100) + "% of the tasks");
return true;
}
@Override
public void setup(Context context) {
this.context = context;
classIndex = ClassIndex.getInstance();
Configuration conf = context.getConfiguration();
float threshold = conf.getFloat(Util.CONF_MINIMUM_DF_OF_HOSTS, 0);
try {
Path headerPath = new Path(conf.get(Util.CONF_HEADER_PATH));
LOG.info("Reading dataset header..." + headerPath.toString());
header = new DataSetHeader(conf, headerPath);
if (!classIndex.isPopulated()) {
classIndex.init(conf);
classIndex.populateIndex();
}
if (threshold > 0.0) {
LOG.info("loading DF values");
hostsWithMinimumDF = Util.getHostsWithDocumentFrequencies(conf, threshold);
}
LOG.info("loading training data...");
loadTrainingInstances(conf.get(Util.CONF_TRAINING_DATE));
} catch (Exception e) {
LOG.error("setup failed with an exception!");
e.printStackTrace();
setupFailedException = e;
}
}
@Override
public void initialize(VertexManagerPluginContext context) {
Configuration conf;
try {
conf = TezUtils.createConfFromUserPayload(context.getUserPayload());
} catch (IOException e) {
throw new TezUncheckedException(e);
}
this.context = context;
this.slowStartMinSrcCompletionFraction =
conf.getFloat(
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_MIN_SRC_FRACTION,
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_MIN_SRC_FRACTION_DEFAULT);
this.slowStartMaxSrcCompletionFraction =
conf.getFloat(
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_MAX_SRC_FRACTION,
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_MAX_SRC_FRACTION_DEFAULT);
if (slowStartMinSrcCompletionFraction < 0
|| slowStartMaxSrcCompletionFraction < slowStartMinSrcCompletionFraction) {
throw new IllegalArgumentException(
"Invalid values for slowStartMinSrcCompletionFraction"
+ "/slowStartMaxSrcCompletionFraction. Min cannot be < 0 and "
+ "max cannot be < min.");
}
enableAutoParallelism =
conf.getBoolean(
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_ENABLE_AUTO_PARALLEL,
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_ENABLE_AUTO_PARALLEL_DEFAULT);
desiredTaskInputDataSize =
conf.getLong(
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_DESIRED_TASK_INPUT_SIZE,
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_DESIRED_TASK_INPUT_SIZE_DEFAULT);
minTaskParallelism =
conf.getInt(
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_MIN_TASK_PARALLELISM,
ShuffleVertexManager.TEZ_AM_SHUFFLE_VERTEX_MANAGER_MIN_TASK_PARALLELISM_DEFAULT);
LOG.info(
"Shuffle Vertex Manager: settings"
+ " minFrac:"
+ slowStartMinSrcCompletionFraction
+ " maxFrac:"
+ slowStartMaxSrcCompletionFraction
+ " auto:"
+ enableAutoParallelism
+ " desiredTaskIput:"
+ desiredTaskInputDataSize
+ " minTasks:"
+ minTaskParallelism);
Map<String, EdgeProperty> inputs = context.getInputVertexEdgeProperties();
for (Map.Entry<String, EdgeProperty> entry : inputs.entrySet()) {
if (entry.getValue().getDataMovementType() == DataMovementType.SCATTER_GATHER) {
String vertex = entry.getKey();
bipartiteSources.put(vertex, new HashSet<Integer>());
}
}
if (bipartiteSources.isEmpty()) {
throw new TezUncheckedException("Atleast 1 bipartite source should exist");
}
// dont track the source tasks here since those tasks may themselves be
// dynamically changed as the DAG progresses.
}
@Override
public synchronized void setConf(Configuration conf) {
super.setConf(conf);
padFraction = conf.getFloat("mapred.jobtracker.taskalloc.capacitypad", 0.01f);
this.eagerTaskInitializationListener = new EagerTaskInitializationListener(conf);
}
public static float getFloat(String name) {
Configuration conf = Services.get().getConf();
return conf.getFloat(name, ConfigUtils.FLOAT_DEFAULT);
}
@Override
public void initialize() {
Configuration conf;
try {
conf = TezUtils.createConfFromUserPayload(getContext().getUserPayload());
} catch (IOException e) {
throw new TezUncheckedException(e);
}
this.slowStartMinSrcCompletionFraction =
conf.getFloat(
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_MIN_SRC_FRACTION,
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_MIN_SRC_FRACTION_DEFAULT);
float defaultSlowStartMaxSrcFraction =
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_MAX_SRC_FRACTION_DEFAULT;
if (slowStartMinSrcCompletionFraction > defaultSlowStartMaxSrcFraction) {
defaultSlowStartMaxSrcFraction = slowStartMinSrcCompletionFraction;
}
this.slowStartMaxSrcCompletionFraction =
conf.getFloat(
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_MAX_SRC_FRACTION,
defaultSlowStartMaxSrcFraction);
if (slowStartMinSrcCompletionFraction < 0
|| slowStartMaxSrcCompletionFraction > 1
|| slowStartMaxSrcCompletionFraction < slowStartMinSrcCompletionFraction) {
throw new IllegalArgumentException(
"Invalid values for slowStartMinSrcCompletionFraction"
+ "/slowStartMaxSrcCompletionFraction. Min cannot be < 0, max cannot be > 1,"
+ " and max cannot be < min.");
}
enableAutoParallelism =
conf.getBoolean(
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_ENABLE_AUTO_PARALLEL,
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_ENABLE_AUTO_PARALLEL_DEFAULT);
desiredTaskInputDataSize =
conf.getLong(
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_DESIRED_TASK_INPUT_SIZE,
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_DESIRED_TASK_INPUT_SIZE_DEFAULT);
minTaskParallelism =
Math.max(
1,
conf.getInt(
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_MIN_TASK_PARALLELISM,
ShuffleVertexManager.TEZ_SHUFFLE_VERTEX_MANAGER_MIN_TASK_PARALLELISM_DEFAULT));
LOG.info(
"Shuffle Vertex Manager: settings"
+ " minFrac:"
+ slowStartMinSrcCompletionFraction
+ " maxFrac:"
+ slowStartMaxSrcCompletionFraction
+ " auto:"
+ enableAutoParallelism
+ " desiredTaskIput:"
+ desiredTaskInputDataSize
+ " minTasks:"
+ minTaskParallelism);
updatePendingTasks();
if (enableAutoParallelism) {
getContext().vertexReconfigurationPlanned();
}
// dont track the source tasks here since those tasks may themselves be
// dynamically changed as the DAG progresses.
}
/**
* Create the memory manager.
*
* @param conf use the configuration to find the maximum size of the memory pool.
*/
MemoryManager(Configuration conf) {
HiveConf.ConfVars poolVar = HiveConf.ConfVars.HIVE_ORC_FILE_MEMORY_POOL;
double maxLoad = conf.getFloat(poolVar.varname, poolVar.defaultFloatVal);
totalMemoryPool =
Math.round(ManagementFactory.getMemoryMXBean().getHeapMemoryUsage().getMax() * maxLoad);
}
