Question

我有一个HBase（v0.94.7）表，其中包含一个列族，并且随着时间的推移会添加列。这些列被命名为它们创建的时间戳，因此除非我查询该行，否则我不知道它具有的所有列。

现在给定一行，我希望原子地删除此列族的所有现有列，并添加一组新的列和值。

所以我想过使用HBase的RowMutations之类的：

RowMutations mutations = new RowMutations(row);

//delete the column family
Delete delete = new Delete(row);
delete.deleteFamily(cf);

//add new columns
Put put = new Put(row);
put.add(cf, col1, v1);
put.add(cf, col2, v2);

//delete column family and add new columns to same family
mutations.add(delete);
mutations.add(put);

table.mutateRow(mutations);

但是这段代码最终做的只是删除列族，它不会添加新列。预期会出现这种情况吗？

如果是这样，那么我怎样才能实现原子用一组新列替换列族的所有列的目标？

以下是相同的测试用例：

import junit.framework.Assert;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.HColumnDescriptor;
import org.apache.hadoop.hbase.HTableDescriptor;
import org.apache.hadoop.hbase.TableExistsException;
import org.apache.hadoop.hbase.client.*;
import org.apache.hadoop.hbase.util.Bytes;
import org.junit.Before;
import org.junit.BeforeClass;
import org.junit.Test;

import java.util.NavigableMap;

public class TestHBaseRowMutations {
    static String tableName = "nnn";
    static byte[] cf1 = Bytes.toBytes("cf1");
    static byte[] row = Bytes.toBytes("r1");
    static HTablePool hTablePool;

    @BeforeClass
    public static void beforeClass() throws Exception {
        Configuration config = HBaseConfiguration.create();
        hTablePool = new HTablePool(config, Integer.MAX_VALUE);
        HBaseAdmin admin = new HBaseAdmin(config);
        HTableDescriptor tableDescriptor = new HTableDescriptor(tableName);
        tableDescriptor.addFamily(new HColumnDescriptor(cf1));
        try {
            admin.createTable(tableDescriptor);
        } catch (TableExistsException ignored){}
    }

    @Before
    public void before() throws Exception {
        HTableInterface table = hTablePool.getTable(tableName);
        try {
            Delete delete = new Delete(row);
            table.delete(delete);
            System.out.println("deleted old row");

            Put put = new Put(row);
            put.add(cf1, Bytes.toBytes("c1"), Bytes.toBytes("v1"));
            put.add(cf1, Bytes.toBytes("c11"), Bytes.toBytes("v11"));
            table.put(put);
            System.out.println("Created row with seed data");
        } finally {
            table.close();
        }
    }


    @Test
    public void testColumnFamilyDeleteRM() throws Exception {
        HTableInterface table = hTablePool.getTable(tableName);
        try {
            RowMutations rm =new RowMutations(row);

            //delete column family cf1
            Delete delete = new Delete(row);
            delete.deleteFamily(cf1);
            rm.add(delete);
            System.out.println("Added delete of cf1 column family to row mutation");

            //add new columns to same column family cf1
            Put put = new Put(row);
            put.add(cf1, Bytes.toBytes("c1"), Bytes.toBytes("new_v1"));
            put.add(cf1, Bytes.toBytes("c11"), Bytes.toBytes("new_v11"));
            rm.add(put);
            System.out.println("Added puts of cf1 column family to row mutation");

            //atomic mutate the row
            table.mutateRow(rm);
            System.out.println("Mutated row");

            //now read the column family cf1 back
            Result result = table.get(new Get(row));
            NavigableMap<byte[], byte[]> familyMap = result.getFamilyMap(cf1);

            //column family cf1 should have 2 columns because of the Put above
            //------Following assert fails as cf1 does not exist anymore, why does cf1 not exist anymore?-------
            Assert.assertNotNull(familyMap);
            Assert.assertEquals(2, familyMap.size());
        } finally {
            table.close();
        }
    }
}

Answer 1

在HBase用户论坛上发布了相同的问题，结果证明这是HBase中的一个错误。

预期的行为是，如果RowMutation对某个列族/列/行有一个删除后跟一个Put-to column-family / column / row，那么Put也应该被尊重（但事实并非如此）目前）。

HBase用户组讨论： http://apache-hbase.679495.n3.nabble.com/Using-RowMutations-to-replace-all-columns-of-a-row-td4045247.html

HBase JIRA同样： https://issues.apache.org/jira/browse/HBASE-8626也提供补丁。

Answer 2

最接近的是将Put上的时间戳设置为高于Delete：

long now = System.currentTimeMillis();

Delete delete = new Delete(row);
delete.deleteFamily(cf1, now);

Put put = new Put(row);
put.add(cf1, col1, now + 1);

RowMutations mutations = new RowMutations(row);
mutations.add(delete);
mutations.add(put);

table.mutateRow(mutations);

可悲的是，这确实意味着现在时间戳为get＆＃39;该列系列中没有任何内容。 Source

Answer 3

有一个要共享的场景，当我们尝试执行RowMutations列表时，每个列表可能包含有效的ROW1：CF1：Q1：V1的Put以及ROW1：CF2：Q1：V1的Delete作为hbase批量操作，如下所示错误

java.lang.RuntimeException：java.lang.UnsupportedOperationException：多调用中没有RowMutations；在使用mutateRow org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithoutRetries（RpcRetryingCaller.java:218）在 org.apache.hadoop.hbase.client.AsyncProcess $ AsyncRequestFutureImpl $ SingleServerRequestRunnable.run（AsyncProcess.java:748）在 java.util.concurrent.Executors $ RunnableAdapter.call（Executors.java:511）在java.util.concurrent.FutureTask.run（FutureTask.java:266）在 java.util.concurrent.ThreadPoolExecutor.runWorker（ThreadPoolExecutor.java:1149）在 java.util.concurrent.ThreadPoolExecutor $ Worker.run（ThreadPoolExecutor.java:624）在java.lang.Thread.run（Thread.java:748）

为解决此问题，我们选择分别执行每个rowMutation。有任何建议，欢迎您。

HBase RowMutations替换行的所有列

3 个答案: