org.apache.hadoop.hbase.mapreduce.TableInputFormatBase Maven / Gradle / Ivy
Show all versions of hbase-mapreduce Show documentation
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hadoop.hbase.mapreduce;
import java.io.Closeable;
import java.io.IOException;
import java.net.InetAddress;
import java.net.InetSocketAddress;
import java.net.UnknownHostException;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import org.apache.hadoop.hbase.HConstants;
import org.apache.hadoop.hbase.HRegionLocation;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.client.Admin;
import org.apache.hadoop.hbase.client.Connection;
import org.apache.hadoop.hbase.client.RegionLocator;
import org.apache.hadoop.hbase.client.Result;
import org.apache.hadoop.hbase.client.Scan;
import org.apache.hadoop.hbase.client.Table;
import org.apache.hadoop.hbase.exceptions.IllegalArgumentIOException;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.hbase.util.Addressing;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.hbase.util.Pair;
import org.apache.hadoop.hbase.util.Strings;
import org.apache.hadoop.mapreduce.InputFormat;
import org.apache.hadoop.mapreduce.InputSplit;
import org.apache.hadoop.mapreduce.JobContext;
import org.apache.hadoop.mapreduce.RecordReader;
import org.apache.hadoop.mapreduce.TaskAttemptContext;
import org.apache.hadoop.net.DNS;
import org.apache.hadoop.util.StringUtils;
import org.apache.yetus.audience.InterfaceAudience;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
/**
* A base for {@link TableInputFormat}s. Receives a {@link Connection}, a {@link TableName}, an
* {@link Scan} instance that defines the input columns etc. Subclasses may use other
* TableRecordReader implementations. Subclasses MUST ensure initializeTable(Connection, TableName)
* is called for an instance to function properly. Each of the entry points to this class used by
* the MapReduce framework, {@link #createRecordReader(InputSplit, TaskAttemptContext)} and
* {@link #getSplits(JobContext)}, will call {@link #initialize(JobContext)} as a convenient
* centralized location to handle retrieving the necessary configuration information. If your
* subclass overrides either of these methods, either call the parent version or call initialize
* yourself.
*
* An example of a subclass:
*
*
* class ExampleTIF extends TableInputFormatBase {
*
* {@literal @}Override
* protected void initialize(JobContext context) throws IOException {
* // We are responsible for the lifecycle of this connection until we hand it over in
* // initializeTable.
* Connection connection = ConnectionFactory.createConnection(HBaseConfiguration.create(
* job.getConfiguration()));
* TableName tableName = TableName.valueOf("exampleTable");
* // mandatory. once passed here, TableInputFormatBase will handle closing the connection.
* initializeTable(connection, tableName);
* byte[][] inputColumns = new byte [][] { Bytes.toBytes("columnA"),
* Bytes.toBytes("columnB") };
* // optional, by default we'll get everything for the table.
* Scan scan = new Scan();
* for (byte[] family : inputColumns) {
* scan.addFamily(family);
* }
* Filter exampleFilter = new RowFilter(CompareOp.EQUAL, new RegexStringComparator("aa.*"));
* scan.setFilter(exampleFilter);
* setScan(scan);
* }
* }
*
*
* The number of InputSplits(mappers) match the number of regions in a table by default. Set
* "hbase.mapreduce.tableinput.mappers.per.region" to specify how many mappers per region, set this
* property will disable autobalance below.\ Set "hbase.mapreduce.tif.input.autobalance" to enable
* autobalance, hbase will assign mappers based on average region size; For regions, whose size
* larger than average region size may assigned more mappers, and for smaller one, they may group
* together to use one mapper. If actual average region size is too big, like 50G, it is not good to
* only assign 1 mapper for those large regions. Use "hbase.mapreduce.tif.ave.regionsize" to set max
* average region size when enable "autobalanece", default mas average region size is 8G.
*/
@InterfaceAudience.Public
public abstract class TableInputFormatBase extends InputFormat {
private static final Logger LOG = LoggerFactory.getLogger(TableInputFormatBase.class);
private static final String NOT_INITIALIZED = "The input format instance has not been properly "
+ "initialized. Ensure you call initializeTable either in your constructor or initialize "
+ "method";
private static final String INITIALIZATION_ERROR = "Cannot create a record reader because of a"
+ " previous error. Please look at the previous logs lines from"
+ " the task's full log for more details.";
/** Specify if we enable auto-balance to set number of mappers in M/R jobs. */
public static final String MAPREDUCE_INPUT_AUTOBALANCE = "hbase.mapreduce.tif.input.autobalance";
/**
* In auto-balance, we split input by ave region size, if calculated region size is too big, we
* can set it.
*/
public static final String MAX_AVERAGE_REGION_SIZE = "hbase.mapreduce.tif.ave.regionsize";
/** Set the number of Mappers for each region, all regions have same number of Mappers */
public static final String NUM_MAPPERS_PER_REGION =
"hbase.mapreduce.tableinput.mappers.per.region";
/**
* Holds the details for the internal scanner.
* @see Scan
*/
private Scan scan = null;
/** The {@link Admin}. */
private Admin admin;
/** The {@link Table} to scan. */
private Table table;
/** The {@link RegionLocator} of the table. */
private RegionLocator regionLocator;
/** The reader scanning the table, can be a custom one. */
private TableRecordReader tableRecordReader = null;
/** The underlying {@link Connection} of the table. */
private Connection connection;
/** Used to generate splits based on region size. */
private RegionSizeCalculator regionSizeCalculator;
/** The reverse DNS lookup cache mapping: IPAddress => HostName */
private HashMap reverseDNSCacheMap = new HashMap<>();
/**
* Builds a {@link TableRecordReader}. If no {@link TableRecordReader} was provided, uses the
* default.
* @param split The split to work with.
* @param context The current context.
* @return The newly created record reader.
* @throws IOException When creating the reader fails.
* @see org.apache.hadoop.mapreduce.InputFormat#createRecordReader(
* org.apache.hadoop.mapreduce.InputSplit, org.apache.hadoop.mapreduce.TaskAttemptContext)
*/
@Override
public RecordReader createRecordReader(InputSplit split,
TaskAttemptContext context) throws IOException {
// Just in case a subclass is relying on JobConfigurable magic.
if (table == null) {
initialize(context);
}
// null check in case our child overrides getTable to not throw.
try {
if (getTable() == null) {
// initialize() must not have been implemented in the subclass.
throw new IOException(INITIALIZATION_ERROR);
}
} catch (IllegalStateException exception) {
throw new IOException(INITIALIZATION_ERROR, exception);
}
TableSplit tSplit = (TableSplit) split;
LOG.info("Input split length: " + StringUtils.humanReadableInt(tSplit.getLength()) + " bytes.");
final TableRecordReader trr =
this.tableRecordReader != null ? this.tableRecordReader : new TableRecordReader();
Scan sc = new Scan(this.scan);
sc.setStartRow(tSplit.getStartRow());
sc.setStopRow(tSplit.getEndRow());
trr.setScan(sc);
trr.setTable(getTable());
return new RecordReader() {
@Override
public void close() throws IOException {
trr.close();
closeTable();
}
@Override
public ImmutableBytesWritable getCurrentKey() throws IOException, InterruptedException {
return trr.getCurrentKey();
}
@Override
public Result getCurrentValue() throws IOException, InterruptedException {
return trr.getCurrentValue();
}
@Override
public float getProgress() throws IOException, InterruptedException {
return trr.getProgress();
}
@Override
public void initialize(InputSplit inputsplit, TaskAttemptContext context)
throws IOException, InterruptedException {
trr.initialize(inputsplit, context);
}
@Override
public boolean nextKeyValue() throws IOException, InterruptedException {
return trr.nextKeyValue();
}
};
}
protected Pair getStartEndKeys() throws IOException {
return getRegionLocator().getStartEndKeys();
}
/**
* Calculates the splits that will serve as input for the map tasks.
* @param context The current job context.
* @return The list of input splits.
* @throws IOException When creating the list of splits fails.
* @see org.apache.hadoop.mapreduce.InputFormat#getSplits( org.apache.hadoop.mapreduce.JobContext)
*/
@Override
public List getSplits(JobContext context) throws IOException {
boolean closeOnFinish = false;
// Just in case a subclass is relying on JobConfigurable magic.
if (table == null) {
initialize(context);
closeOnFinish = true;
}
// null check in case our child overrides getTable to not throw.
try {
if (getTable() == null) {
// initialize() must not have been implemented in the subclass.
throw new IOException(INITIALIZATION_ERROR);
}
} catch (IllegalStateException exception) {
throw new IOException(INITIALIZATION_ERROR, exception);
}
try {
List splits = oneInputSplitPerRegion();
// set same number of mappers for each region
if (context.getConfiguration().get(NUM_MAPPERS_PER_REGION) != null) {
int nSplitsPerRegion = context.getConfiguration().getInt(NUM_MAPPERS_PER_REGION, 1);
List res = new ArrayList<>();
for (int i = 0; i < splits.size(); i++) {
List tmp = createNInputSplitsUniform(splits.get(i), nSplitsPerRegion);
res.addAll(tmp);
}
return res;
}
// The default value of "hbase.mapreduce.input.autobalance" is false.
if (context.getConfiguration().getBoolean(MAPREDUCE_INPUT_AUTOBALANCE, false)) {
long maxAveRegionSize =
context.getConfiguration().getLong(MAX_AVERAGE_REGION_SIZE, 8L * 1073741824); // 8GB
return calculateAutoBalancedSplits(splits, maxAveRegionSize);
}
// return one mapper per region
return splits;
} finally {
if (closeOnFinish) {
closeTable();
}
}
}
/**
* Create one InputSplit per region
* @return The list of InputSplit for all the regions
* @throws IOException throws IOException
*/
private List oneInputSplitPerRegion() throws IOException {
if (regionSizeCalculator == null) {
// Initialize here rather than with the other resources because this involves
// a full scan of meta, which can be heavy. We might as well only do it if/when necessary.
regionSizeCalculator = createRegionSizeCalculator(getRegionLocator(), getAdmin());
}
TableName tableName = getTable().getName();
Pair keys = getStartEndKeys();
if (keys == null || keys.getFirst() == null || keys.getFirst().length == 0) {
HRegionLocation regLoc =
getRegionLocator().getRegionLocation(HConstants.EMPTY_BYTE_ARRAY, false);
if (null == regLoc) {
throw new IOException("Expecting at least one region.");
}
List splits = new ArrayList<>(1);
long regionSize = regionSizeCalculator.getRegionSize(regLoc.getRegion().getRegionName());
// In the table input format for single table we do not need to
// store the scan object in table split because it can be memory intensive and redundant
// information to what is already stored in conf SCAN. See HBASE-25212
TableSplit split =
new TableSplit(tableName, null, HConstants.EMPTY_BYTE_ARRAY, HConstants.EMPTY_BYTE_ARRAY,
regLoc.getHostnamePort().split(Addressing.HOSTNAME_PORT_SEPARATOR)[0], regionSize);
splits.add(split);
return splits;
}
List splits = new ArrayList<>(keys.getFirst().length);
for (int i = 0; i < keys.getFirst().length; i++) {
if (!includeRegionInSplit(keys.getFirst()[i], keys.getSecond()[i])) {
continue;
}
byte[] startRow = scan.getStartRow();
byte[] stopRow = scan.getStopRow();
// determine if the given start an stop key fall into the region
if (
(startRow.length == 0 || keys.getSecond()[i].length == 0
|| Bytes.compareTo(startRow, keys.getSecond()[i]) < 0)
&& (stopRow.length == 0 || Bytes.compareTo(stopRow, keys.getFirst()[i]) > 0)
) {
byte[] splitStart =
startRow.length == 0 || Bytes.compareTo(keys.getFirst()[i], startRow) >= 0
? keys.getFirst()[i]
: startRow;
byte[] splitStop =
(stopRow.length == 0 || Bytes.compareTo(keys.getSecond()[i], stopRow) <= 0)
&& keys.getSecond()[i].length > 0 ? keys.getSecond()[i] : stopRow;
HRegionLocation location = getRegionLocator().getRegionLocation(keys.getFirst()[i], false);
// The below InetSocketAddress creation does a name resolution.
InetSocketAddress isa = new InetSocketAddress(location.getHostname(), location.getPort());
if (isa.isUnresolved()) {
LOG.warn("Failed resolve " + isa);
}
InetAddress regionAddress = isa.getAddress();
String regionLocation;
regionLocation = reverseDNS(regionAddress);
byte[] regionName = location.getRegion().getRegionName();
String encodedRegionName = location.getRegion().getEncodedName();
long regionSize = regionSizeCalculator.getRegionSize(regionName);
// In the table input format for single table we do not need to
// store the scan object in table split because it can be memory intensive and redundant
// information to what is already stored in conf SCAN. See HBASE-25212
TableSplit split = new TableSplit(tableName, null, splitStart, splitStop, regionLocation,
encodedRegionName, regionSize);
splits.add(split);
if (LOG.isDebugEnabled()) {
LOG.debug("getSplits: split -> " + i + " -> " + split);
}
}
}
return splits;
}
/**
* Create n splits for one InputSplit, For now only support uniform distribution
* @param split A TableSplit corresponding to a range of rowkeys
* @param n Number of ranges after splitting. Pass 1 means no split for the range Pass 2 if
* you want to split the range in two;
* @return A list of TableSplit, the size of the list is {@code n}
*/
protected List createNInputSplitsUniform(InputSplit split, int n)
throws IllegalArgumentIOException {
if (split == null || !(split instanceof TableSplit)) {
throw new IllegalArgumentIOException(
"InputSplit for CreateNSplitsPerRegion can not be null + "
+ "and should be instance of TableSplit");
}
// if n < 1, then still continue using n = 1
n = n < 1 ? 1 : n;
List res = new ArrayList<>(n);
if (n == 1) {
res.add(split);
return res;
}
// Collect Region related information
TableSplit ts = (TableSplit) split;
TableName tableName = ts.getTable();
String regionLocation = ts.getRegionLocation();
String encodedRegionName = ts.getEncodedRegionName();
long regionSize = ts.getLength();
byte[] startRow = ts.getStartRow();
byte[] endRow = ts.getEndRow();
// For special case: startRow or endRow is empty
if (startRow.length == 0 && endRow.length == 0) {
startRow = new byte[1];
endRow = new byte[1];
startRow[0] = 0;
endRow[0] = -1;
}
if (startRow.length == 0 && endRow.length != 0) {
startRow = new byte[1];
startRow[0] = 0;
}
if (startRow.length != 0 && endRow.length == 0) {
endRow = new byte[startRow.length];
for (int k = 0; k < startRow.length; k++) {
endRow[k] = -1;
}
}
// Split Region into n chunks evenly
byte[][] splitKeys = Bytes.split(startRow, endRow, true, n - 1);
for (int i = 0; i < splitKeys.length - 1; i++) {
// In the table input format for single table we do not need to
// store the scan object in table split because it can be memory intensive and redundant
// information to what is already stored in conf SCAN. See HBASE-25212
// notice that the regionSize parameter may be not very accurate
TableSplit tsplit = new TableSplit(tableName, null, splitKeys[i], splitKeys[i + 1],
regionLocation, encodedRegionName, regionSize / n);
res.add(tsplit);
}
return res;
}
/**
* Calculates the number of MapReduce input splits for the map tasks. The number of MapReduce
* input splits depends on the average region size. Make it 'public' for testing
* @param splits The list of input splits before balance.
* @param maxAverageRegionSize max Average region size for one mapper
* @return The list of input splits.
* @throws IOException When creating the list of splits fails.
* @see org.apache.hadoop.mapreduce.InputFormat#getSplits( org.apache.hadoop.mapreduce.JobContext)
*/
public List calculateAutoBalancedSplits(List splits,
long maxAverageRegionSize) throws IOException {
if (splits.size() == 0) {
return splits;
}
List resultList = new ArrayList<>();
long totalRegionSize = 0;
for (int i = 0; i < splits.size(); i++) {
TableSplit ts = (TableSplit) splits.get(i);
totalRegionSize += ts.getLength();
}
long averageRegionSize = totalRegionSize / splits.size();
// totalRegionSize might be overflow, and the averageRegionSize must be positive.
if (averageRegionSize <= 0) {
LOG.warn("The averageRegionSize is not positive: " + averageRegionSize + ", "
+ "set it to Long.MAX_VALUE " + splits.size());
averageRegionSize = Long.MAX_VALUE / splits.size();
}
// if averageRegionSize is too big, change it to default as 1 GB,
if (averageRegionSize > maxAverageRegionSize) {
averageRegionSize = maxAverageRegionSize;
}
// if averageRegionSize is too small, we do not need to allocate more mappers for those 'large'
// region
// set default as 16M = (default hdfs block size) / 4;
if (averageRegionSize < 16 * 1048576) {
return splits;
}
for (int i = 0; i < splits.size(); i++) {
TableSplit ts = (TableSplit) splits.get(i);
TableName tableName = ts.getTable();
String regionLocation = ts.getRegionLocation();
String encodedRegionName = ts.getEncodedRegionName();
long regionSize = ts.getLength();
if (regionSize >= averageRegionSize) {
// make this region as multiple MapReduce input split.
int n =
(int) Math.round(Math.log(((double) regionSize) / ((double) averageRegionSize)) + 1.0);
List temp = createNInputSplitsUniform(ts, n);
resultList.addAll(temp);
} else {
// if the total size of several small continuous regions less than the average region size,
// combine them into one MapReduce input split.
long totalSize = regionSize;
byte[] splitStartKey = ts.getStartRow();
byte[] splitEndKey = ts.getEndRow();
int j = i + 1;
while (j < splits.size()) {
TableSplit nextRegion = (TableSplit) splits.get(j);
long nextRegionSize = nextRegion.getLength();
if (
totalSize + nextRegionSize <= averageRegionSize
&& Bytes.equals(splitEndKey, nextRegion.getStartRow())
) {
totalSize = totalSize + nextRegionSize;
splitEndKey = nextRegion.getEndRow();
j++;
} else {
break;
}
}
i = j - 1;
// In the table input format for single table we do not need to
// store the scan object in table split because it can be memory intensive and redundant
// information to what is already stored in conf SCAN. See HBASE-25212
TableSplit t = new TableSplit(tableName, null, splitStartKey, splitEndKey, regionLocation,
encodedRegionName, totalSize);
resultList.add(t);
}
}
return resultList;
}
String reverseDNS(InetAddress ipAddress) throws UnknownHostException {
String hostName = this.reverseDNSCacheMap.get(ipAddress);
if (hostName == null) {
String ipAddressString = null;
try {
ipAddressString = DNS.reverseDns(ipAddress, null);
} catch (Exception e) {
// We can use InetAddress in case the jndi failed to pull up the reverse DNS entry from the
// name service. Also, in case of ipv6, we need to use the InetAddress since resolving
// reverse DNS using jndi doesn't work well with ipv6 addresses.
ipAddressString = InetAddress.getByName(ipAddress.getHostAddress()).getHostName();
}
if (ipAddressString == null) {
throw new UnknownHostException("No host found for " + ipAddress);
}
hostName = Strings.domainNamePointerToHostName(ipAddressString);
this.reverseDNSCacheMap.put(ipAddress, hostName);
}
return hostName;
}
/**
* Test if the given region is to be included in the InputSplit while splitting the regions of a
* table.
*
* This optimization is effective when there is a specific reasoning to exclude an entire region
* from the M-R job, (and hence, not contributing to the InputSplit), given the start and end keys
* of the same.
* Useful when we need to remember the last-processed top record and revisit the [last, current)
* interval for M-R processing, continuously. In addition to reducing InputSplits, reduces the
* load on the region server as well, due to the ordering of the keys.
*
* Note: It is possible that endKey.length() == 0
, for the last (recent) region.
*
* Override this method, if you want to bulk exclude regions altogether from M-R. By default, no
* region is excluded( i.e. all regions are included).
* @param startKey Start key of the region
* @param endKey End key of the region
* @return true, if this region needs to be included as part of the input (default).
*/
protected boolean includeRegionInSplit(final byte[] startKey, final byte[] endKey) {
return true;
}
/**
* Allows subclasses to get the {@link RegionLocator}.
*/
protected RegionLocator getRegionLocator() {
if (regionLocator == null) {
throw new IllegalStateException(NOT_INITIALIZED);
}
return regionLocator;
}
/**
* Allows subclasses to get the {@link Table}.
*/
protected Table getTable() {
if (table == null) {
throw new IllegalStateException(NOT_INITIALIZED);
}
return table;
}
/**
* Allows subclasses to get the {@link Admin}.
*/
protected Admin getAdmin() {
if (admin == null) {
throw new IllegalStateException(NOT_INITIALIZED);
}
return admin;
}
/**
* Allows subclasses to initialize the table information.
* @param connection The Connection to the HBase cluster. MUST be unmanaged. We will close.
* @param tableName The {@link TableName} of the table to process.
*/
protected void initializeTable(Connection connection, TableName tableName) throws IOException {
if (this.table != null || this.connection != null) {
LOG.warn("initializeTable called multiple times. Overwriting connection and table "
+ "reference; TableInputFormatBase will not close these old references when done.");
}
this.table = connection.getTable(tableName);
this.regionLocator = connection.getRegionLocator(tableName);
this.admin = connection.getAdmin();
this.connection = connection;
this.regionSizeCalculator = null;
}
@InterfaceAudience.Private
protected RegionSizeCalculator createRegionSizeCalculator(RegionLocator locator, Admin admin)
throws IOException {
return new RegionSizeCalculator(locator, admin);
}
/**
* Gets the scan defining the actual details like columns etc.
* @return The internal scan instance.
*/
public Scan getScan() {
if (this.scan == null) this.scan = new Scan();
return scan;
}
/**
* Sets the scan defining the actual details like columns etc.
* @param scan The scan to set.
*/
public void setScan(Scan scan) {
this.scan = scan;
}
/**
* Allows subclasses to set the {@link TableRecordReader}.
* @param tableRecordReader A different {@link TableRecordReader} implementation.
*/
protected void setTableRecordReader(TableRecordReader tableRecordReader) {
this.tableRecordReader = tableRecordReader;
}
/**
* Handle subclass specific set up. Each of the entry points used by the MapReduce framework,
* {@link #createRecordReader(InputSplit, TaskAttemptContext)} and {@link #getSplits(JobContext)},
* will call {@link #initialize(JobContext)} as a convenient centralized location to handle
* retrieving the necessary configuration information and calling
* {@link #initializeTable(Connection, TableName)}. Subclasses should implement their initialize
* call such that it is safe to call multiple times. The current TableInputFormatBase
* implementation relies on a non-null table reference to decide if an initialize call is needed,
* but this behavior may change in the future. In particular, it is critical that initializeTable
* not be called multiple times since this will leak Connection instances.
*/
protected void initialize(JobContext context) throws IOException {
}
/**
* Close the Table and related objects that were initialized via
* {@link #initializeTable(Connection, TableName)}.
*/
protected void closeTable() throws IOException {
close(admin, table, regionLocator, connection);
admin = null;
table = null;
regionLocator = null;
connection = null;
regionSizeCalculator = null;
}
private void close(Closeable... closables) throws IOException {
for (Closeable c : closables) {
if (c != null) {
c.close();
}
}
}
}