Hive：简单查询不启用Mapreduce job而启用Fetch task

存储，学习，共享

如果你想查询某个表的某一列，Hive默认是会启用MapReduce Job来完成这个任务，如下：

01	hive> SELECT id, money FROM m limit 10;

02	Total MapReduce jobs = 1

03	Launching Job 1 out of 1

04	Number of reduce tasks is set to 0 since there's no reduce operator

05	Cannot run job locally: Input Size (= 235105473) is larger than

06	hive.exec.mode.local.auto.inputbytes.max (= 134217728)

07	Starting Job = job_1384246387966_0229, Tracking URL =

08

09	http://l-datalogm1.data.cn1:9981/proxy/application_1384246387966_0229/

10

11	Kill Command = /home/q/hadoop-2.2.0/bin/hadoop job

12	-kill job_1384246387966_0229

13	hadoop job information for Stage-1: number of mappers: 1;

14	number of reducers: 0

15	2013-11-13 11:35:16,167 Stage-1 map = 0%,  reduce = 0%

16	2013-11-13 11:35:21,327 Stage-1 map = 100%,  reduce = 0%,

17	Cumulative CPU 1.26 sec

18	2013-11-13 11:35:22,377 Stage-1 map = 100%,  reduce = 0%,

19	Cumulative CPU 1.26 sec

20	MapReduce Total cumulative CPU time: 1 seconds 260 msec

21	Ended Job = job_1384246387966_0229

22	MapReduce Jobs Launched:

23	Job 0: Map: 1   Cumulative CPU: 1.26 sec

24	HDFS Read: 8388865 HDFS Write: 60 SUCCESS

25	Total MapReduce CPU Time Spent: 1 seconds 260 msec

26

OK

27	1       122

28	1       185

29	1       231

30	1       292

31	1       316

32	1       329

33	1       355

34	1       356

35	1       362

36	1       364

37	Time taken: 16.802 seconds, Fetched: 10 row(s)

　　我们都知道，启用MapReduce Job是会消耗系统开销的。对于这个问题，从Hive0.10.0版本开始，对于简单的不需要聚合的类似SELECT <col> from <table> LIMIT n语句，不需要起MapReduce job，直接通过Fetch task获取数据，可以通过下面几种方法实现：
　　方法一：

01	hive> set hive.fetch.task.conversion=more;

02	hive> SELECT id, money FROM m limit 10;

03

OK

04	1       122

05	1       185

06	1       231

07	1       292

08	1       316

09	1       329

10	1       355

11	1       356

12	1       362

13	1       364

14	Time taken: 0.138 seconds, Fetched: 10 row(s)

上面 set hive.fetch.task.conversion=more;开启了Fetch任务，所以对于上述简单的列查询不在启用MapReduce job！
　　方法二：

1	bin/hive --hiveconf hive.fetch.task.conversion=more

　　方法三：
上面的两种方法都可以开启了Fetch任务，但是都是临时起作用的；如果你想一直启用这个功能，可以在${HIVE_HOME}/conf/hive-site.xml里面加入以下配置：

01	<property>

02	<name>hive.fetch.task.conversion</name>

03	<value>more</value>

04	<description>

05	Some select queries can be converted to single FETCH task

06	minimizing latency.Currently the query should be single

07	sourced not having any subquery and should not have

08	any aggregations or distincts (which incurrs RS),

09	lateral views and joins.

10	1. minimal : SELECT STAR, FILTER on partition columns, LIMIT only

11	2. more    : SELECT, FILTER, LIMIT only (+TABLESAMPLE, virtual columns)

12	</description>

13	</property>

这样就可以长期启用Fetch任务了，很不错吧，也赶紧去试试吧！ 转载自过往记忆（http://www.iteblog.com/）
本文链接地址: 《Hive：简单查询不启用Mapreduce job而启用Fetch task》（http://www.iteblog.com/archives/831）