Flume-ng将数据插入HBase-0.96.0

Flume-ng将数据插入HBase-0.96.0

这篇介绍过flume数据插入hdfs和普通目录（），本文继续介绍flume-ng将数据插入hbase-0.96.0.

首先，修改node中flume文件夹下conf目录中的flume-node.conf文件（原配置参考上文），对其做如下修改：

1. agent.sinks = k1
3. agent.sinks.k1.type = hbase
4. agent.sinks.k1.table = hello
5. agent.sinks.k1.columnFamily = cf
6. agent.sinks.k1.column = col1
7. agent.sinks.k1.serializer = org.apache.flume.sink.hbase.SimpleHbaseEventSerializer
8. agent.sinks.k1.channel = memoryChannel

agent.sinks = k1

agent.sinks.k1.type = hbase
agent.sinks.k1.table = hello
agent.sinks.k1.columnFamily = cf
agent.sinks.k1.column = col1
agent.sinks.k1.serializer = org.apache.flume.sink.hbase.SimpleHbaseEventSerializer
agent.sinks.k1.channel = memoryChannel

不过和上文不同的是，这次要想得到成功结果就没那么简单了，由于依赖的版本问题。此处需要将flume的lib文件夹下的protobuf用Hadoop-2.2.0中的2.5.0版本替换，还需要用hadoop-2.2.0中的guava替换flume的lib文件夹下的guava，删除原来相应的jar文件。启动即可生效。

flume-ng里面的SimpleHbaseEventSerializer只提供了最简单的数据插入hbase功能，如果还有其他需要，就得自己写HbaseEventSerializer类，在apache-flume-1.4.0-src/flume-ng-sinks/flume-ng-hbase-sink/src/main/java中定义自己的类，实现flume中的HbaseEventSerializer接口。一个简单的实例如下：

1. publicclass MyHBaseSerializer implements HbaseEventSerializer {
3. privatestaticfinal String[] COLUMNS = "column1,column2".split(",");
4. privatestaticfinal String[] PARAMS = "col1,col2".split(",");
6. privatebyte[] columnFamily = "cf".getBytes();
7. privatebyte[] content;
10. @Override
11. publicvoid configure(Context context) {
13. }
15. @Override
16. publicvoid configure(ComponentConfiguration conf) {
18. }
20. @Override
21. publicvoid initialize(Event event, byte[] columnFamily) {
22. this.content = event.getBody();
23. this.columnFamily = columnFamily;
24. }
26. @Override
27. public List<Row> getActions() {
28. String string = Bytes.toString(content);
29. String value1 = string.substring(0,string.length()/2);
30. String value2 = string.substring(string.length()/2, string.length());
32. Map<String,String> map = Maps.newHashMap();
33. map.put(PARAMS[0], value1);
34. map.put(PARAMS[1], value2);
37. List<Row> actions = new LinkedList<Row>();
38. String rowKey = String.valueOf(System.currentTimeMillis());
40. Put put = new Put(Bytes.toBytes(rowKey));
41. for (int i = 0; i < COLUMNS.length; i++) {
42. String value = map.get(PARAMS[i]);
43. if (value == null)
44. value = "";
45. put.add(columnFamily, Bytes.toBytes(COLUMNS[i]), Bytes.toBytes(value));
46. }
47. actions.add(put);
48. return actions;
49. }
51. @Override
52. public List<Increment> getIncrements() {
53. List<Increment> incs = new LinkedList<Increment>();
54. return incs;
55. }
57. @Override
58. publicvoid close() {
60. }
62. }

public class MyHBaseSerializer implements HbaseEventSerializer {

    private static final String[] COLUMNS = "column1,column2".split(",");
    private static final String[] PARAMS = "col1,col2".split(",");

    private byte[] columnFamily = "cf".getBytes();
    private byte[] content;


    @Override
    public void configure(Context context) {

    }

    @Override
    public void configure(ComponentConfiguration conf) {

    }

    @Override
    public void initialize(Event event, byte[] columnFamily) {
        this.content = event.getBody();
        this.columnFamily = columnFamily;
    }

    @Override
    public List<Row> getActions() {
        String string = Bytes.toString(content);
        String value1 = string.substring(0,string.length()/2);
        String value2 = string.substring(string.length()/2, string.length());

        Map<String,String> map = Maps.newHashMap();
        map.put(PARAMS[0], value1);
        map.put(PARAMS[1], value2);


        List<Row> actions = new LinkedList<Row>();
        String rowKey = String.valueOf(System.currentTimeMillis());

        Put put = new Put(Bytes.toBytes(rowKey));
        for (int i = 0; i < COLUMNS.length; i++) {
            String value = map.get(PARAMS[i]);
            if (value == null)
                value = "";
            put.add(columnFamily, Bytes.toBytes(COLUMNS[i]), Bytes.toBytes(value));
        }
        actions.add(put);
        return actions;
    }

    @Override
    public List<Increment> getIncrements() {
        List<Increment> incs = new LinkedList<Increment>();
        return incs;
    }

    @Override
    public void close() {

    }

}

该类实现的功能是将文件中的内容按行切分程两部分，分别插入列名为column1和column2的两列中，rowKey为当前时间。完成后将flume-ng代码重新编译打包。然后将flume-ng目录里面的lib文件夹的相应的jar文件替换。然后将上文中的agent.sinks.k1.serializer 值改为test..MyHBaseSerializer即可。其中test为包名。

HBase 的详细介绍：请点这里
HBase 的下载地址：请点这里

Hadoop+HBase搭建云存储总结 PDF

HBase 结点之间时间不一致造成regionserver启动失败

Hadoop+ZooKeeper+HBase集群配置

Hadoop集群安装&HBase实验环境搭建

基于Hadoop集群的HBase集群的配置 ‘

Hadoop安装部署笔记之-HBase完全分布模式安装

单机版搭建HBase环境图文教程详解