Hcatalog是apache开源的对于表和底层数据管理统一服务平台,目前最新release版本是0.5,不过需要hive 0.10支持,由于我们hive集群版本是0.9.0,所以只能降级使用hcatalog 0.4,由于hcatalog中所有的底层数据信息都是保存在hive metastore里,所以hive版本升级后schema变动或者api变动会对hacatalog产生影响,因此在hive 0.11中已经集成了了hcatalog,以后也会成为hive的一部分,而不是独立的项目。
HCatalog底层依赖于Hive Metastore,执行过程中会创建一个HiveMetaStoreClient,通过这个instance提供的api来获取表结构数据,如果是local metastore mode的话,会直接返回一个HiveMetaStore.HMSHandler,如果是remote mode的话(hive.metastore.local设置为false),会依据hive.metastore.uris(比如thrift://10.1.8.42:9083, thrift://10.1.8.51:9083)中设定的一串uri逐一顺序建立连接。只要有一个链接建立就可以了,同时为了避免所有client都和第一个uri建立连接,导致负载过大,我加了点小trick,对这串uris随机shuffle来做load balance
由于我们的集群开启了kerberos security,需要获取DelegationToken,但是local mode是不支持的,所以只用能remote mode
HiveMetaStoreClient.java
public String getDelegationToken(String owner, String renewerKerberosPrincipalName) throws
MetaException, TException {
if (localMetaStore) {
throw new UnsupportedOperationException("getDelegationToken() can be " +
"called only in thrift (non local) mode");
}
return client.get_delegation_token(owner, renewerKerberosPrincipalName);
}
HCatInputFormat和HCatOutputFormat提供一些mapreduce api来读取表和写入表
HCatInputFormat API:
public static void setInput(Job job,
InputJobInfo inputJobInfo) throws IOException;
先实例化一个InputJobInfo对象,该对象包含三个参数dbname,tablename,filter,然后传给setInput函数,来读取相应的数据
public static HCatSchema getTableSchema(JobContext context)
throws IOException;
在运行时(比如mapper阶段的setup函数中),可以传进去JobContext,调用静态getTableSchema来获取先前setInput时设置的table schema信息
HCatOutputFormat API:
public static void setOutput(Job job, OutputJobInfo outputJobInfo) throws IOException;
OutPutJobInfo接受三个参数databaseName, tableName, partitionValues,其中第三个参数类型是Map<String, String>,partition key放在map key里,partition value放在对应map key的value中,该参数可传入null或空map,如果指定的partition存在的话,会抛org.apache.hcatalog.common.HCatException : 2002 : Partition already present with given partition key values
比如要要写入指定的partition(dt=’2013-06-13′,country=’china’ ),可以这样写
Map<String, String> partitionValues = new HashMap<String, String>();
partitionValues.put("dt", "2013-06-13");
partitionValues.put("country", "china");
HCatTableInfo info = HCatTableInfo.getOutputTableInfo(dbName, tblName, partitionValues);
HCatOutputFormat.setOutput(job, info);
public static HCatSchema getTableSchema(JobContext context) throws IOException;
获取之前HCatOutputFormat.setOutput指定的table schema信息
public static void setSchema(final Job job, final HCatSchema schema) throws IOException;
设置最终写入数据的schema信息,若不调用这个方法,则默认会使用table schema信息
下面提供一个完整mapreduce例子计算一天每个guid访问页面次数,map阶段从表中读取guid字段,reduce阶段统计该guid对应pageview的总数,然后写回另外一张带有guid和count字段的表中
import java.io.IOException;
import java.util.Iterator;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.conf.Configured;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.io.WritableComparable;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.Reducer;
import org.apache.hadoop.util.Tool;
import org.apache.hadoop.util.ToolRunner;
import org.apache.hcatalog.data.DefaultHCatRecord;
import org.apache.hcatalog.data.HCatRecord;
import org.apache.hcatalog.data.schema.HCatSchema;
import org.apache.hcatalog.mapreduce.HCatInputFormat;
import org.apache.hcatalog.mapreduce.HCatOutputFormat;
import org.apache.hcatalog.mapreduce.InputJobInfo;
import org.apache.hcatalog.mapreduce.OutputJobInfo;
public class GroupByGuid extends Configured implements Tool {
@SuppressWarnings("rawtypes")
public static class Map extends
Mapper<WritableComparable, HCatRecord, Text, IntWritable> {
HCatSchema schema;
Text guid;
IntWritable one;
@Override
protected void setup(org.apache.hadoop.mapreduce.Mapper.Context context)
throws IOException, InterruptedException {
guid = new Text();
one = new IntWritable(1);
schema = HCatInputFormat.getTableSchema(context);
}
@Override
protected void map(WritableComparable key, HCatRecord value,
Context context) throws IOException, InterruptedException {
guid.set(value.getString("guid", schema));
context.write(guid, one);
}
}
@SuppressWarnings("rawtypes")
public static class Reduce extends
Reducer<Text, IntWritable, WritableComparable, HCatRecord> {
HCatSchema schema;
@Override
protected void setup(org.apache.hadoop.mapreduce.Reducer.Context context)
throws IOException, InterruptedException {
schema = HCatOutputFormat.getTableSchema(context);
}
@Override
protected void reduce(Text key, Iterable<IntWritable> values,
Context context) throws IOException, InterruptedException {
int sum = 0;
Iterator<IntWritable> iter = values.iterator();
while (iter.hasNext()) {
sum++;
iter.next();
}
HCatRecord record = new DefaultHCatRecord(2);
record.setString("guid", schema, key.toString());
record.setInteger("count", schema, sum);
context.write(null, record);
}
}
@Override
public int run(String[] args) throws Exception {
Configuration conf = getConf();
String dbname = args[0];
String inputTable = args[1];
String filter = args[2];
String outputTable = args[3];
int reduceNum = Integer.parseInt(args[4]);
Job job = new Job(conf,
"GroupByGuid, Calculating every guid's pageview");
HCatInputFormat.setInput(job,
InputJobInfo.create(dbname, inputTable, filter));
job.setJarByClass(GroupByGuid.class);
job.setInputFormatClass(HCatInputFormat.class);
job.setMapperClass(Map.class);
job.setReducerClass(Reduce.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(IntWritable.class);
job.setOutputKeyClass(WritableComparable.class);
job.setOutputValueClass(DefaultHCatRecord.class);
job.setNumReduceTasks(reduceNum);
HCatOutputFormat.setOutput(job,
OutputJobInfo.create(dbname, outputTable, null));
HCatSchema s = HCatOutputFormat.getTableSchema(job);
HCatOutputFormat.setSchema(job, s);
job.setOutputFormatClass(HCatOutputFormat.class);
return (job.waitForCompletion(true) ? 0 : 1);
}
public static void main(String[] args) throws Exception {
int exitCode = ToolRunner.run(new GroupByGuid(), args);
System.exit(exitCode);
}
}
其实hcatalog还支持动态分区dynamic partition,我们可以在OutJobInfo中指定部分partition keyvalue pair,在运行时候根据传进来的值设置HCatRecord对应的其他partition keyvalue pair,这样就能在一个job中同时写多个partition了
本文链接http://blog.csdn.net/lalaguozhe/article/details/9083905,转载请注明