Hbase手动split、compact、merge、批量合并分区

hbase shell命令手动Split:
// 手动 split region
split 'TABLENAME'
split 'REGIONNAME'
split 'ENCODED_REGIONNAME'
split 'TABLENAME','splitKey'
split 'REGIONNAME','splitKey'
split 'ENCODED_REGIONNAME','splitKey'
hbase shell命令合并:
// 手动 merge region
// 合并相邻的两个Region
merge_region 'ENCODED_REGIONNAME', 'ENCODED_REGIONNAME'
// 强制合并两个Region
merge_region 'ENCODED_REGIONNAME', 'ENCODED_REGIONNAME', true
批量合并分区 merge_regions (java代码参考)
import java.io.IOException;
import java.net.URI;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
 
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.HRegionInfo;
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.ConnectionFactory;
import org.apache.hadoop.hbase.util.Bytes;
 
/**
 * 使用hbase java api对hbase的特定表按条件进行合并,由于未找到合适的api,所以hbase的region大小从hdfs上获取,如果有java api能实现该功能希望大家告知
 * 阀值:lower_size:小region,直接合并,默认100M
 *     upper_size:大region,按条件合并,默认5G
 *     region_max_size:最大region大小,建议和region split的阀值大小相同,默认10G
 *     least_region_count:最少region个数,默认10个
 * 算法:首先从hdfs的hbase存储路径获得table的region名称和size对应关系的map,然后通过api获取排序好的region实例List,然后遍历该List进行两两合并
 * 规则:(待合并的region定义为A,B两个region)
 * 当前region大小小于upper_size:
 * 如果A region为空,则将当前region赋值给A
 * 如果A region不为空,则将当前region赋值给B,则直接调用api对A,B进行合并,成功之后清空A,B继续遍历
 * 当前region大小大于upper_size:
 * 如果A region为空,放弃处理该region,直接继续遍历
 * 如果A region不为空,并且A的大小小于lower_size或者A+当前的大小小于region_max_size,则将当前region赋值给B,则直接调用api对A,B进行合并,成功之后清空A,B继续遍历
 * @author taochy
 * 
 *
 */
public class HBaseRegionMergeUtil {
	
	//小region,直接合并,默认100M
	private static long lower_size = 100 * 1024 * 1024L;
	//大region,按条件合并,默认5G
	private static long upper_size = 5 * 1024 * 1024 * 1024L;
	//最大region大小,建议和region split的阀值大小相同,默认10G
	private static long region_max_size = 10 * 1024 * 1024 * 1024L;
	//最少region个数,默认10个
	private static int least_region_count = 10;
	
	
	public static void main(String[] args) {
		Configuration conf = HBaseConfiguration.create();
		conf.set("hbase.rootdir", "hdfs://hb1:9000/hbase");
		conf.set("hbase.master", "hb1:60000");
		conf.set("hbase.zookeeper.property.clientPort", "2181");
		conf.set("hbase.zookeeper.quorum",
				"hb1:2181,hb2:2181,hb3:2181,hd4:2181,hd5:2181,hd6:2181,hd7:2181,hd8:2181,hd9:2181,hd10:2181,hd11:2181");
		mergeRegion4Table(conf, "staticlog_2018");
	}
	
	/**
	 * 按表进行region的merge
	 * @param conf
	 * @param tableName
	 */
	public static void mergeRegion4Table(Configuration conf, String tableName) {
 
		FileSystem fs = null;
		conf.set("fs.defaultFS", "hdfs://hb1:9000");
		conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
		Map<String, Long> tableRegionSizeMap = null;
		long currentTimeMillis = 0L;
		long currentTimeMillis2 = 0L;
		int merge_count = 0;
 
		try {
			fs = FileSystem.get(URI.create("hdfs://hb1:9000"), conf);
			tableRegionSizeMap = getTableRegionSizeMap(fs, tableName);
		} catch (IOException e) {
			e.printStackTrace();
		}
 
		try {
			Connection connection = ConnectionFactory.createConnection(conf);
			Admin admin = connection.getAdmin();
			List<HRegionInfo> tableRegions = admin.getTableRegions(TableName.valueOf(tableName));
			System.out.println("table " + tableName + " has " + tableRegions.size() + " regions!");
			// 合并的执行条件,大于最小的region个数
			if (tableRegions.size() > least_region_count) {
				byte[] regionA = null;
				long sizeA = 0L;
				byte[] regionB = null;
				for (HRegionInfo info : tableRegions) {
					String regionName = info.getRegionNameAsString();
					regionName = regionName.substring(regionName.lastIndexOf(".") - 32, regionName.lastIndexOf("."));
					Long size = tableRegionSizeMap.get(regionName);
					if (size != 0L) {
						// region大小小于upper_size
						if (size < upper_size) {
							if (regionA == null) {
								//如果A region为空,则将当前region赋值给A
								regionA = info.getRegionName();
								sizeA = size;
							} else {
								//如果A region不为空,则将当前region赋值给B,则直接调用api对A,B进行合并,成功之后清空A,B继续遍历
								regionB = info.getRegionName();
								currentTimeMillis = System.currentTimeMillis();
								admin.mergeRegions(regionA, regionB, true);
								currentTimeMillis2 = System.currentTimeMillis();
								System.out.println(Bytes.toString(regionA) + " & " + Bytes.toString(regionB)
										+ "merge cost " + (currentTimeMillis2 - currentTimeMillis) + " milliseconds!");
								regionA = null;
								regionB = null;
								sizeA = 0L;
								merge_count++;
							}
						} else {
							// region大小大于upper_size
							if (regionA == null) {
								//如果A region为空,放弃处理该region,直接继续遍历
								continue;
							} else {
								//如果A region不为空,并且A的大小小于lower_size或者A+当前的大小小于region_max_size,则将当前region赋值给B,则直接调用api对A,B进行合并,成功之后清空A,B继续遍历
								if (sizeA < lower_size || (sizeA + size) < region_max_size) {
									regionB = info.getRegionName();
									currentTimeMillis = System.currentTimeMillis();
									admin.mergeRegions(regionA, regionB, true);
									currentTimeMillis2 = System.currentTimeMillis();
									System.out.println(
											Bytes.toString(regionA) + " & " + Bytes.toString(regionB) + "merge cost "
													+ (currentTimeMillis2 - currentTimeMillis) + " milliseconds!");
									regionA = null;
									regionB = null;
									sizeA = 0L;
									merge_count++;
								} else {
									regionA = null;
									sizeA = 0L;
								}
							}
						}
					} else {
						System.out.println(regionName + "is empty,ignore!");
					}
				}
				admin.close();
			}
		} catch (IOException e) {
			// TODO Auto-generated catch block
			e.printStackTrace();
		}
		System.out.println("merge run " + merge_count + " times,continue!");
		// if(merge_count != 0){
		// System.out.println("merge run "+ merge_count +" times,continue!");
		// //递归操作,直到不需要继续合并
		// mergeRegion4Table(conf,tableName);
		// }else {
		// System.out.println("no need to merge any more,finish!");
		// }
	}
 
	/**
	 * 获得table的region名称和size对应关系的map
	 * @param fs
	 * @param tableName
	 * @return
	 */
	public static Map<String, Long> getTableRegionSizeMap(FileSystem fs, String tableName) {
		Map<String, Long> tableRegionSizeMap = new HashMap<>();
		String hbase_path = "/hbase/data/";
		if (tableName.contains(":")) {
			String[] split = tableName.split(":");
			hbase_path = hbase_path + split[0] + "/" + split[1];
		} else {
			hbase_path = hbase_path + "default/" + tableName;
		}
		Path table_path = new Path(hbase_path);
		try {
			FileStatus[] files = fs.listStatus(table_path);
			System.out.println(files.length);
			for (FileStatus file : files) {
				String name = file.getPath().getName();
				long length = fs.getContentSummary(file.getPath()).getLength();
				tableRegionSizeMap.put(name, length);
			}
		} catch (IOException e) {
			e.printStackTrace();
		}
		return tableRegionSizeMap;
	}
}


补充

//表压缩 :
major_compact
//压缩整个区域(region)(可单独压缩列、列族) :
major_compact ‘table’
//刷新表(将mem中缓存刷入Hfile) :
flush ‘table’

通过以上手动Split方法可临时性解决Region倾斜问题

Logo

为开发者提供学习成长、分享交流、生态实践、资源工具等服务,帮助开发者快速成长。

更多推荐