Cassandra杂记14-谈谈Cassandra的客户端

最近试用了一段时间Cassandra，将Oracle中的数据导入进来，遇到了问题然后解决问题，收获挺大。在这个过程中，除了设计一个合理的数据模型，再就是使用Cassandra API进行交互了。

Cassandra在设计的时候，就是支持Thrift的，这意味着我们可以使用多种语言开发。

对于Cassandra的开发本身而言，这是使用Thrift的好处：支持多语言。坏处也是显而易见的：Thrift API功能过于简单，不具备在生产环境使用的条件。

在Cassandra Wiki页面上，也有基于Thrift API开发的更加高级的API，各个语言都有，具体信息可以参考：http://wiki.apache.org/cassandra/ClientExamples。

这次只谈谈下面两类Java的客户端：

1 Thrift Java API

2 hector

Thrift Java API

这个是Cassandra自带的最简单的一类API，这个文件在apache-cassandra-0.5.1.jar中包含了。可以直接使用。我们也可以自己安装一个Thrift，然后通过cassandra.thrift文件自动生成。

如果你要使用Cassandra，那么我们必须要了解Thrift API，毕竟所有的其他更加高级的API都是基于这个来包装的。

提供的功能

插入数据

插入数据需要指定keyspace，ColumnFamily， Column，Key，Value，timestamp和数据同步级别。（如何需要了Cassandra的解数据模型，可以参考《大话Cassandra数据模型》）

/**

 * Insert a Column consisting of (column_path.column, value, timestamp) at the given column_path.column_family and optional

 * column_path.super_column. Note that column_path.column is here required, since a SuperColumn cannot directly contain binary

 * values -- it can only contain sub-Columns.

 * 

 * @param keyspace

 * @param key

 * @param column_path

 * @param value

 * @param timestamp

 * @param consistency_level

 */

public void insert(String keyspace, String key, ColumnPath column_path, byte[] value, long timestamp, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

/**

 * Insert Columns or SuperColumns across different Column Families for the same row key. batch_mutation is a

 * map<string, list<ColumnOrSuperColumn>> -- a map which pairs column family names with the relevant ColumnOrSuperColumn

 * objects to insert.

 * 

 * @param keyspace

 * @param key

 * @param cfmap

 * @param consistency_level

 */

public void batch_insert(String keyspace, String key, Map<String,List<ColumnOrSuperColumn>> cfmap, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

读取数据

获取一个查询条件精确的值。

/**

 * Get the Column or SuperColumn at the given column_path. If no value is present, NotFoundException is thrown. (This is

 * the only method that can throw an exception under non-failure conditions.)

 * 

 * @param keyspace

 * @param key

 * @param column_path

 * @param consistency_level

 */

public ColumnOrSuperColumn get(String keyspace, String key, ColumnPath column_path, int consistency_level) throws InvalidRequestException, NotFoundException, UnavailableException, TimedOutException, TException;

/**

 * Perform a get for column_path in parallel on the given list<string> keys. The return value maps keys to the

 * ColumnOrSuperColumn found. If no value corresponding to a key is present, the key will still be in the map, but both

 * the column and super_column references of the ColumnOrSuperColumn object it maps to will be null.

 * 

 * @param keyspace

 * @param keys

 * @param column_path

 * @param consistency_level

 */

public Map<String,ColumnOrSuperColumn> multiget(String keyspace, List<String> keys, ColumnPath column_path, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

获取某一个keyspace，Key，ColumnFamily，SuperColumn（如果有的话需要指定）下面的相关数据：只查询Column的name符合条件的相关数据（SlicePredicate）。

/**

 * Get the group of columns contained by column_parent (either a ColumnFamily name or a ColumnFamily/SuperColumn name

 * pair) specified by the given SlicePredicate. If no matching values are found, an empty list is returned.

 * 

 * @param keyspace

 * @param key

 * @param column_parent

 * @param predicate

 * @param consistency_level

 */

public List<ColumnOrSuperColumn> get_slice(String keyspace, String key, ColumnParent column_parent, SlicePredicate predicate, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

/**

 * Performs a get_slice for column_parent and predicate for the given keys in parallel.

 * 

 * @param keyspace

 * @param keys

 * @param column_parent

 * @param predicate

 * @param consistency_level

 */

public Map<String,List<ColumnOrSuperColumn>> multiget_slice(String keyspace, List<String> keys, ColumnParent column_parent, SlicePredicate predicate, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

查询Key的取值范围（使用这个功能需要使用order-preserving partitioner）。

/**

 * @deprecated; use get_range_slice instead

 * 

 * @param keyspace

 * @param column_family

 * @param start

 * @param finish

 * @param count

 * @param consistency_level

 */

public List<String> get_key_range(String keyspace, String column_family, String start, String finish, int count, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

/**

 * returns a subset of columns for a range of keys.

 * 

 * @param keyspace

 * @param column_parent

 * @param predicate

 * @param start_key

 * @param finish_key

 * @param row_count

 * @param consistency_level

 */

public List<KeySlice> get_range_slice(String keyspace, ColumnParent column_parent, SlicePredicate predicate, String start_key, String finish_key, int row_count, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

查询系统的信息。

/**

 * get property whose value is of type string.

 * 

 * @param property

 */

public String get_string_property(String property) throws TException;

/**

 * get property whose value is list of strings.

 * 

 * @param property

 */

public List<String> get_string_list_property(String property) throws TException;

/**

 * describe specified keyspace

 * 

 * @param keyspace

 */

public Map<String,Map<String,String>> describe_keyspace(String keyspace) throws NotFoundException, TException;

通过这些操作，我们可以了解到系统的信息。

其中一个比较有意思的查询信息是：token map，通过这个我们可以知道哪些Cassandra Service是可以提供服务的。

删除数据

/**

 * Remove data from the row specified by key at the granularity specified by column_path, and the given timestamp. Note

 * that all the values in column_path besides column_path.column_family are truly optional: you can remove the entire

 * row by just specifying the ColumnFamily, or you can remove a SuperColumn or a single Column by specifying those levels too.

 * 

 * @param keyspace

 * @param key

 * @param column_path

 * @param timestamp

 * @param consistency_level

 */

public void remove(String keyspace, String key, ColumnPath column_path, long timestamp, int consistency_level) throws InvalidRequestException, UnavailableException, TimedOutException, TException;

这里需要注意的是，由于一致性的问题。这里的删除操作不会立即删除所有机器上的该数据，但是最终会一致。

程序范例

import java.util.List;

import java.io.UnsupportedEncodingException;

import org.apache.thrift.transport.TTransport;

import org.apache.thrift.transport.TSocket;

import org.apache.thrift.protocol.TProtocol;

import org.apache.thrift.protocol.TBinaryProtocol;

import org.apache.thrift.TException;

import org.apache.cassandra.service.*;

public class CClient

{

    public static void main(String[] args)

    throws TException, InvalidRequestException, UnavailableException, UnsupportedEncodingException, NotFoundException

    {

        TTransport tr = new TSocket("localhost", 9160);

        TProtocol proto = new TBinaryProtocol(tr);

        Cassandra.Client client = new Cassandra.Client(proto);

        tr.open();

        String key_user_id = "逖靖寒的世界";

        // insert data

        long timestamp = System.currentTimeMillis();

        client.insert("Keyspace1",

                      key_user_id,

                      new ColumnPath("Standard1", null, "网址".getBytes("UTF-8")),

                      "http://gpcuster.cnblogs.com".getBytes("UTF-8"),

                      timestamp,

                      ConsistencyLevel.ONE);

        client.insert("Keyspace1",

                      key_user_id,

                      new ColumnPath("Standard1", null, "作者".getBytes("UTF-8")),

                      "逖靖寒".getBytes("UTF-8"),

                      timestamp,

                      ConsistencyLevel.ONE);

        // read single column

        ColumnPath path = new ColumnPath("Standard1", null, "name".getBytes("UTF-8"));

        System.out.println(client.get("Keyspace1", key_user_id, path, ConsistencyLevel.ONE));

        // read entire row

        SlicePredicate predicate = new SlicePredicate(null, new SliceRange(new byte[0], new byte[0], false, 10));

        ColumnParent parent = new ColumnParent("Standard1", null);

        List<ColumnOrSuperColumn> results = client.get_slice("Keyspace1", key_user_id, parent, predicate, ConsistencyLevel.ONE);

        for (ColumnOrSuperColumn result : results)

        {

            Column column = result.column;

            System.out.println(new String(column.name, "UTF-8") + " -> " + new String(column.value, "UTF-8"));

        }

        tr.close();

    }

}

优点与缺点

优点：简单高效

缺点：功能简单，无法提供连接池，错误处理等功能，不适合直接在生产环境使用。

Hector

Hector是基于Thrift Java API包装的一个Java客户端，提供一个更加高级的一个抽象。

程序范例

package me.prettyprint.cassandra.service;

import static me.prettyprint.cassandra.utils.StringUtils.bytes;

import static me.prettyprint.cassandra.utils.StringUtils.string;

import org.apache.cassandra.service.Column;

import org.apache.cassandra.service.ColumnPath;

public class ExampleClient {

  public static void main(String[] args) throws IllegalStateException, PoolExhaustedException,

      Exception {

    CassandraClientPool pool = CassandraClientPoolFactory.INSTANCE.get();

    CassandraClient client = pool.borrowClient("localhost", 9160);

    // A load balanced version would look like this:

    // CassandraClient client = pool.borrowClient(new String[] {"cas1:9160", "cas2:9160", "cas3:9160"});

    try {

      Keyspace keyspace = client.getKeyspace("Keyspace1");

      ColumnPath columnPath = new ColumnPath("Standard1", null, bytes("网址"));

      // insert

      keyspace.insert("逖靖寒的世界", columnPath, bytes("http://gpcuster.cnblogs.com"));

      // read

      Column col = keyspace.getColumn("逖靖寒的世界", columnPath);

      System.out.println("Read from cassandra: " + string(col.getValue()));

    } finally {

      // return client to pool. do it in a finally block to make sure it's executed

      pool.releaseClient(client);

    }

  }

}

优点

1 提供连接池。

2 提供错误处理：当操作失败的时候，Hector会根据系统信息（token map）自动连接另一个Cassandra Service。

3 编程接口容易使用。

4 支持JMX。

缺点

1 不支持多线程的环境。

2 keyspace封装过多（数据校验和数据重新封装），如果进行大量的数据操作，这里的消耗需要考虑。

3 错误处理不够人性化：如果所有的Cassandra Service都非常繁忙，那么经过多次操作失败后，最终的结果失败。

总结

Hector已经是一个基本足够使用的Java客户端了，但是还是缺乏一些相关的功能，比如：

1 线程安全。

2 支持自动的多线程查询和插入，提高操作效率。

3 人性化的错误处理机制。

4 避免过多的封装。