在Flink中我们有时候需要分析数据1点到2点的范围,可是经过Region又比较慢,这时候我们就可以定制
TableInputFormat
来实现我们的需求了,我们还可以采用Flink的DataSet
的方式读取,另外下面还有Spark
读取的例子。
使用教程
Md5Util.java
import org.apache.commons.codec.binary.Hex;
import java.security.MessageDigest;
import java.security.NoSuchAlgorithmException;
public class Md5Util {
public static String md5(byte[] key) {
return md5(key, 0, key.length);
}
public static String md5(byte[] key, int offset, int length) {
try {
MessageDigest e = MessageDigest.getInstance("MD5");
e.update(key, offset, length);
byte[] digest = e.digest();
return new String(Hex.encodeHex(digest));
} catch (NoSuchAlgorithmException var5) {
throw new RuntimeException("Error computing MD5 hash", var5);
}
}
public static String md5(String str) {
return md5(str.getBytes());
}
public static String md5(String str,int offset, int length) {
return md5(str.getBytes(),offset,length);
}
}
数据Split
方式
private Connection connection;
private Admin admin;
@Before
public void init() throws Exception {
System.setProperty("java.security.krb5.conf", "/etc/krb5.conf");
System.setProperty("sun.security.krb5.debug", "false");
final String user = "hbase/abc.demo.com@DEMO.COM";
final String keyPath = "/home/dounine/kerberos/lake.keytab";
Configuration conf = new Configuration();
conf.addResource("hbase-site.xml");
UserGroupInformation.setConfiguration(conf);
UserGroupInformation.loginUserFromKeytab(user, keyPath);
connection = ConnectionFactory.createConnection(conf);
admin = connection.getAdmin();
}
@Test
public void createTable() throws IOException {
TableName table = TableName.valueOf("logTable1");
TableDescriptorBuilder tableDesc = TableDescriptorBuilder.newBuilder(table);
tableDesc.setValue(TableDescriptorBuilder.SPLIT_POLICY,KeyPrefixRegionSplitPolicy.class.getName());
tableDesc.setValue(KeyPrefixRegionSplitPolicy.PREFIX_LENGTH_KEY,"2");
ColumnFamilyDescriptor extCF = ColumnFamilyDescriptorBuilder.newBuilder("ext".getBytes()).build();
ColumnFamilyDescriptor deviceCF = ColumnFamilyDescriptorBuilder.newBuilder("device".getBytes()).build();
ColumnFamilyDescriptor locationCF = ColumnFamilyDescriptorBuilder.newBuilder("location".getBytes()).build();
tableDesc.setColumnFamilies(Arrays.asList(extCF,locationCF,deviceCF));
try {
byte[][] splitKeys = new byte[4][];
splitKeys[0] = Bytes.toBytes("00");
splitKeys[1] = Bytes.toBytes("40");
splitKeys[2] = Bytes.toBytes("80");
splitKeys[3] = Bytes.toBytes("c0");
admin.createTable(tableDesc.build(),splitKeys);
} catch (IOException e) {
e.printStackTrace();
}
}
logTable1
数据写入方式
public class HbaseKerberos{
private static final Logger LOGGER = LoggerFactory.getLogger(HbaseKerberos.class);
private static final DateTimeFormatter dtf = DateTimeFormatter.ofPattern("yyyyMMddHHmmssSSS");
private static final String TABLE_NAME = "logTable1";
public void insertDataToHbase1(String appKey,List<Log> hasDatas) throws IOException {
Table table = HbaseUtils.getTable(TABLE_NAME);
Long sumCount = 0L;
/**
* 常规值
*/
byte[] extCF = Bytes.toBytes("ext");//CF列族
Random random = new Random();
List<Put> rows = new ArrayList<>();
for (Log logEntity : hasDatas) {
JSONObject dataJsonObject = logEntity.getData();
JSONObject extJsonObject = dataJsonObject.getJSONObject("ext");
String userId = extJsonObject.getString("userId");
String timeStr = logEntity.getTime().format(dtf);
String md5Str = Md5Util.md5(userId);
String rowKey = new StringBuilder()
.append(md5Str.substring(0,2))//md5出来的前两位最高为ff,00~ff为256位,后期Region可以增加那么多,足够使用了。
.append("|")
.append(timeStr)//时间
.append("|")
.append(CrcUtil.getCrcValue(appKey))
.append("|")
.append(md5Str.substring(2,8))
.append("|")
.append(Md5Util.md5(UUID.randomUUID().toString()).substring(0,2))
.toString();
Put row = new Put(Bytes.toBytes(rowKey));
for(String keyName : extJsonObject.keySet()){
String value = extJsonObject.getString(keyName);
if(StringUtils.isNotBlank(value)){
row.addColumn(extCF, Bytes.toBytes(keyName), Bytes.toBytes(value));
}
}
row.addColumn(extCF, Bytes.toBytes("time"), Bytes.toBytes(logEntity.getTime().toString()));
/**
* 设备信息
*/
putFieldToRow(logEntity.getData(),"device",row);
/**
* 位置信息
*/
putFieldToRow(logEntity.getData(),"location",row);
rows.add(row);
}
for(Integer[] durtation : LimitUtil.getLimits(rows.size(),1000)){
Object[] results = new Object[(durtation[1]-durtation[0])];
try {
table.batch(rows.subList(durtation[0], durtation[1]),results);
} catch (InterruptedException e) {
e.printStackTrace();
}
sumCount += (durtation[1]-durtation[0]);
}
LOGGER.info("write data count:" + sumCount);
}
}
logTable1
数据
00|20180518203401772|2352356512|4519 column=ext:appKey, timestamp=1533646292389, value=898b7e90-5754-11e8-983c-6b4bcc3b7c2e
f3|f1
00|20180518203401772|2352356512|4519 column=ext:channelCode, timestamp=1533646292389, value=guanlan-resurrection-002-
f3|f1
00|20180518203401772|2352356512|4519 column=ext:createDateTime, timestamp=1533646292389, value=1526646836093
f3|f1
00|20180518203401772|2352356512|4519 column=ext:retain, timestamp=1533646292389, value=17670
f3|f1
00|20180518203401772|2352356512|4519 column=ext:scene, timestamp=1533646292389, value=1007
f3|f1
00|20180518203401772|2352356512|4519 column=ext:shareId, timestamp=1533646292389, value=ogJmG5ItE_nBCS3pg5XCvGotGI1c
f3|f1
00|20180518203401772|2352356512|4519 column=ext:time, timestamp=1533646292389, value=2018-05-18T20:34:01
f3|f1
00|20180518203401772|2352356512|4519 column=ext:type, timestamp=1533646292389, value=login_in
f3|f1
00|20180518203401772|2352356512|4519 column=ext:userId, timestamp=1533646292389, value=ogJmG5KRcIxtyg7UmcRHFCn6YiAQ
f3|f1
00|20180518203406167|2352356512|4519 column=ext:appKey, timestamp=1533646347725, value=898b7e90-5754-11e8-983c-6b4bcc3b7c2e
f3|54
00|20180518203406167|2352356512|4519 column=ext:channelCode, timestamp=1533646347725, value=guanlan-regular-001-
f3|54
00|20180518203406167|2352356512|4519 column=ext:createDateTime, timestamp=1533646347725, value=1526646839075
f3|54
00|20180518203406167|2352356512|4519 column=ext:retain, timestamp=1533646347725, value=17670
f3|54
00|20180518203406167|2352356512|4519 column=ext:shareId, timestamp=1533646347725, value=ogJmG5KRcIxtyg7UmcRHFCn6YiAQ
f3|54
00|20180518203406167|2352356512|4519 column=ext:time, timestamp=1533646347725, value=2018-05-18T20:34:06
f3|54
00|20180518203406167|2352356512|4519 column=ext:type, timestamp=1533646347725, value=sharesuccess
f3|54
00|20180518203406167|2352356512|4519 column=ext:userId, timestamp=1533646347725, value=ogJmG5KRcIxtyg7UmcRHFCn6YiAQ
f3|54
00|20180518203407144|2352356512|5ca1 column=ext:appKey, timestamp=1533646294045, value=898b7e90-5754-11e8-983c-6b4bcc3b7c2e
c4|bc
00|20180518203407144|2352356512|5ca1 column=ext:createDateTime, timestamp=1533646294045, value=1526646849745
c4|bc
00|20180518203407144|2352356512|5ca1 column=ext:retain, timestamp=1533646294045, value=17670
c4|bc
00|20180518203407144|2352356512|5ca1 column=ext:scene, timestamp=1533646294045, value=1037
c4|bc
00|20180518203407144|2352356512|5ca1 column=ext:time, timestamp=1533646294045, value=2018-05-18T20:34:07
c4|bc
00|20180518203407144|2352356512|5ca1 column=ext:type, timestamp=1533646294045, value=login_in
CustomTableInputFormat.java
import org.apache.commons.lang3.StringUtils;
import org.apache.hadoop.hbase.HRegionLocation;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.mapreduce.RegionSizeCalculator;
import org.apache.hadoop.hbase.mapreduce.TableInputFormat;
import org.apache.hadoop.hbase.mapreduce.TableSplit;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.hbase.util.Strings;
import org.apache.hadoop.mapreduce.InputSplit;
import org.apache.hadoop.mapreduce.JobContext;
import org.apache.hadoop.net.DNS;
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;
public class CustomTableInputFormat extends TableInputFormat {
private HashMap<InetAddress, String> reverseDNSCacheMap =
new HashMap<>();
private List<String> keys = new ArrayList<>();
public CustomTableInputFormat(){
super();
for(int i =0;i<256;i++){
keys.add(StringUtils.substring("00"+Integer.toHexString(i),-2));
}
}
@Override
public List<InputSplit> getSplits(JobContext context) throws IOException {
super.initialize(context);
TableName tableName = super.getTable().getName();
RegionSizeCalculator sizeCalculator = new RegionSizeCalculator(getRegionLocator(), getAdmin());
List<InputSplit> splits = new ArrayList<>();
for (String key : keys) {
HRegionLocation location = getRegionLocator().getRegionLocation(Bytes.toBytes(key), false);
InetSocketAddress isa = new InetSocketAddress(location.getHostname(), location.getPort());
InetAddress regionAddress = isa.getAddress();
String regionLocation;
regionLocation = reverseDNS(regionAddress);
byte[] regionName = location.getRegion().getRegionName();
String encodedRegionName = location.getRegion().getEncodedName();
long regionSize = sizeCalculator.getRegionSize(regionName);
byte[] splitStart = Bytes.add(Bytes.toBytes(key+"|"),this.getScan().getStartRow());
byte[] splitStop = Bytes.add(Bytes.toBytes(key+"|"),this.getScan().getStopRow());
TableSplit split = new TableSplit(tableName, this.getScan(),
splitStart, splitStop, regionLocation, encodedRegionName, regionSize);
splits.add(split);
}
return splits;
}
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) {
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;
}
}
Flink例子
static Configuration conf;
static {
HadoopKrbLogin.login();
conf = new Configuration();
String tableName = "logTable1";
conf.addResource("hbase-site.xml");
Scan scan = new Scan();
scan.setCaching(1000);
scan.withStartRow("201805182039".getBytes());
scan.withStopRow("201805182040".getBytes());
scan.setCacheBlocks(false);
conf.set(org.apache.hadoop.hbase.mapreduce.TableInputFormat.INPUT_TABLE, tableName);
ClientProtos.Scan proto = null;
try {
proto = ProtobufUtil.toScan(scan);
} catch (IOException e) {
e.printStackTrace();
}
String ScanToString = Base64.encodeBytes(proto.toByteArray());
conf.set(org.apache.hadoop.hbase.mapreduce.TableInputFormat.SCAN, ScanToString);
}
public static void main(String[] args) throws Exception {
final ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
DataSource<Tuple2<ImmutableBytesWritable, Result>> hbase = env.createInput(
HadoopInputs.createHadoopInput(
new CustomTableInputFormat(),
ImmutableBytesWritable.class,
Result.class,
Job.getInstance(conf)
)
);
DataSet<LogEntity> toTuple = hbase.map(
new MapFunction<Tuple2<ImmutableBytesWritable, Result>, LogEntity>() {
public LogEntity map(Tuple2<ImmutableBytesWritable, Result> record) throws Exception {
Result result = record.f1;
return result2Entity(result);
}
});
}
private static LogEntity result2Entity(Result result) {
JSONObject root = new JSONObject();
JSONObject ext = new JSONObject();
JSONObject device = new JSONObject();
JSONObject location = new JSONObject();
for (Cell cell : result.rawCells()) {
byte[] family = CellUtil.cloneFamily(cell);
byte[] column = CellUtil.cloneQualifier(cell);
byte[] value = CellUtil.cloneValue(cell);
String columnName = Bytes.toString(column);
if ("ext".equals(Bytes.toString(family))) {
if ("durationTime".equals(columnName)) {
ext.put(columnName, Bytes.toLong(value));
} else if ("time".equals(columnName)) {
root.put(columnName, Bytes.toString(value));
root.put("timeLong", DateUtil.getMill(LocalDateTime.parse(Bytes.toString(value))));
} else {
ext.put(columnName, Bytes.toString(value));
}
} else if ("device".equals(Bytes.toString(family))) {
device.put(columnName, Bytes.toString(value));
} else if ("location".equals(Bytes.toString(family))) {
location.put(columnName, Bytes.toString(value));
}
}
JSONObject data = new JSONObject();
if (device.keySet().size() > 0) {
data.put("device", device);
}
if (location.keySet().size() > 0) {
data.put("location", location);
}
data.put("ext", ext);
root.put("data", data);
return JSON.parseObject(root.toString(), LogEntity.class);
}
Spark 例子
public class SimpleApp implements Serializable {
static Configuration cfg = null;
static {
HadoopKrbLogin.login();
cfg = new Configuration();
String tableName = "logTable1";
cfg.addResource("hbase-site.xml");
Scan scan = new Scan();
scan.setCaching(1000);
scan.withStartRow("201805182039".getBytes());
scan.withStopRow("201805182040".getBytes());
scan.setCacheBlocks(false);
cfg.set(TableInputFormat.INPUT_TABLE, tableName);
ClientProtos.Scan proto = null;
try {
proto = ProtobufUtil.toScan(scan);
} catch (IOException e) {
e.printStackTrace();
}
String ScanToString = Base64.encodeBytes(proto.toByteArray());
cfg.set(TableInputFormat.SCAN, ScanToString);
}
public static void main(String[] args) {
SparkConf sparkConf = new SparkConf()
.setMaster("local")
.setAppName("HbaseDemo");
JavaSparkContext jsc = new JavaSparkContext(sparkConf);
JavaPairRDD<ImmutableBytesWritable, Result> hBaseRDD =
jsc.newAPIHadoopRDD(cfg, CustomTableInputFormat.class, ImmutableBytesWritable.class, Result.class);
// do some transformation
JavaRDD<LogEntity> rdd1 = hBaseRDD.mapPartitions((FlatMapFunction<Iterator<Tuple2<ImmutableBytesWritable, Result>>, LogEntity>)
tuple2Iterator -> {
List<LogEntity> logEntities = new ArrayList<>();
while (tuple2Iterator.hasNext()) {
Tuple2<ImmutableBytesWritable, Result> tuple = tuple2Iterator.next();
Result result = tuple._2;
String rowKey = Bytes.toString(result.getRow());
logEntities.add(result2Entity(result));
}
return logEntities.iterator();
});
}
private static LogEntity result2Entity(Result result) {
JSONObject root = new JSONObject();
JSONObject ext = new JSONObject();
JSONObject device = new JSONObject();
JSONObject location = new JSONObject();
for (Cell cell : result.rawCells()) {
byte[] family = CellUtil.cloneFamily(cell);
byte[] column = CellUtil.cloneQualifier(cell);
byte[] value = CellUtil.cloneValue(cell);
String columnName = Bytes.toString(column);
if ("ext".equals(Bytes.toString(family))) {
if ("durationTime".equals(columnName)) {
ext.put(columnName, Bytes.toLong(value));
} else if ("time".equals(columnName)) {
root.put(columnName, Bytes.toString(value));
root.put("timeLong", DateUtil.getMill(LocalDateTime.parse(Bytes.toString(value))));
} else {
ext.put(columnName, Bytes.toString(value));
}
} else if ("device".equals(Bytes.toString(family))) {
device.put(columnName, Bytes.toString(value));
} else if ("location".equals(Bytes.toString(family))) {
location.put(columnName, Bytes.toString(value));
}
}
JSONObject data = new JSONObject();
if (device.keySet().size() > 0) {
data.put("device", device);
}
if (location.keySet().size() > 0) {
data.put("location", location);
}
data.put("ext", ext);
root.put("data", data);
return JSON.parseObject(root.toString(), LogEntity.class);
}
网友评论