使用dbms_parallel_execute来完成DML的并行(r3笔记第1天)

在工作中使用并行可以极大的提高工作效率。可以Object,session.hint级别引入并行。可以使大量的数据处理更加高效。 比如现在有一个表 t 有1000万行，如果想以这个表为基础，把数据选择性的插入另外一个表t2， 使用Insert into t2 select *from t; 使用并行来处理也没有问题，但是如果使用dbms_parallel_execute也是一种很不错的选择。 使用dbms_parallel_execute的实现方式和parallel还有一定的差别。 这个包在11g开始引入，可能初次接触的时候会被它大量的功能所淹没，不知道从何开始。 举个例子来说明一下。 我们创建一个表 t，限于环境的情况，目前做一个百万级别的数据dml操作，使用dbms_parallel_execute来完成。 创建表t.

SQL> drop table t;

Table dropped.

SQL> create table t as select object_id,object_name from dba_objects;

Table created. 创建表t2,我们专门专门多加了一个字段。session_id。到时候会有用处。

SQL> drop table t2;

Table dropped.

SQL> create table t2 as select t1.*,0 session_id from t t1 where 1=0;

Table created. 使用如下的存储过程来模拟一个dml的处理过程。传入的参数，是根据rowid来处理。 create or replace procedure serial(p_lo_rid in rowid,p_hi_rid in rowid) is begin for x in (select object_id object_id,object_name object_name from t where rowid between p_lo_rid and p_hi_rid) loop insert into t2(object_id,object_name,session_id) values(x.object_id,x.object_name,sys_context('userenv','sessionid')); end loop; end; / 使用dbms_parallel_execute来创建一个Job,以1万条数据分单位进行数据的rowid切分。

begin dbms_parallel_execute.create_task('PROCESS TASK'); dbms_parallel_execute.create_chunks_by_rowid ( task_name=>'PROCESS TASK', table_owner=>user, table_name=>'T', by_row=>false, chunk_size=>10000); end; / 通过dba_parallel_execute_chunks可以查看到切分后的rowid情况。

set pages 200 select *from ( select chunk_id,status,start_rowid,end_rowid from dba_parallel_execute_chunks where task_name='PROCESS TASK' order by chunk_id ); 查看切分后的情况，我们可以把切分后的每一个子块称为chunk。可以通过这个语句来简单的监控进度。

CHUNK_ID STATUS START_ROWID END_ROWID ---------- -------------------- ------------------ ------------------ 600 UNASSIGNED AAAEQCAAFAAAACAAAA AAAEQCAAFAAAAD/CcP 601 UNASSIGNED AAAEQCAAFAAAAEAAAA AAAEQCAAFAAAAF/CcP 602 UNASSIGNED AAAEQCAAFAAAAGAAAA AAAEQCAAFAAAAH/CcP 603 UNASSIGNED AAAEQCAAFAAAAIAAAA AAAEQCAAFAAAAJ/CcP 604 UNASSIGNED AAAEQCAAFAAAAKAAAA AAAEQCAAFAAAAL/CcP 605 UNASSIGNED AAAEQCAAFAAAAWAAAA AAAEQCAAFAAAAX/CcP 606 UNASSIGNED AAAEQCAAFAAAAYAAAA AAAEQCAAFAAAAZ/CcP 607 UNASSIGNED AAAEQCAAFAAAAaAAAA AAAEQCAAFAAAAb/CcP 608 UNASSIGNED AAAEQCAAFAAAAcAAAA AAAEQCAAFAAAAd/CcP 609 UNASSIGNED AAAEQCAAFAAAEsYAAA AAAEQCAAFAAAEsfCcP 610 UNASSIGNED AAAEQCAAFAAAEsgAAA AAAEQCAAFAAAEsnCcP 可以使用如下的部分来开始处理数据。启用了4个并行，并行度可以情况来提高。:start_id,:end_id是上面对应的rowid.

begin dbms_parallel_execute.run_task (task_name=>'PROCESS TASK', sql_stmt=>'begin serial(:start_id,:end_id); end;', language_flag=>DBMS_SQL.NATIVE, parallel_level=>4); end; /

select *from ( select chunk_id,status,start_rowid,end_rowid from dba_parallel_execute_chunks where task_name='PROCESS TASK' order by chunk_id ); 数据处理的进度可以查看得到。 CHUNK_ID STATUS START_ROWID END_ROWID ---------- -------------------- ------------------ ------------------ 600 PROCESSED AAAEQCAAFAAAACAAAA AAAEQCAAFAAAAD/CcP 601 PROCESSED AAAEQCAAFAAAAEAAAA AAAEQCAAFAAAAF/CcP 602 PROCESSED AAAEQCAAFAAAAGAAAA AAAEQCAAFAAAAH/CcP 603 PROCESSED AAAEQCAAFAAAAIAAAA AAAEQCAAFAAAAJ/CcP 604 PROCESSED AAAEQCAAFAAAAKAAAA AAAEQCAAFAAAAL/CcP 605 PROCESSED AAAEQCAAFAAAAWAAAA AAAEQCAAFAAAAX/CcP 606 PROCESSED AAAEQCAAFAAAAYAAAA AAAEQCAAFAAAAZ/CcP 607 PROCESSED AAAEQCAAFAAAAaAAAA AAAEQCAAFAAAAb/CcP 608 PROCESSED AAAEQCAAFAAAAcAAAA AAAEQCAAFAAAAd/CcP 处理完数据之后，就可以删除这个job了。

begin dbms_parallel_execute.drop_task('PROCESS TASK'); end; /

我们可以在t2的新增列中看到每个对应的parallel处理的数据情况，可以看到数据的处理还是很平均的。 select session_id,count(*) from t2 group by session_id order by session_id; SESSION_ID COUNT(*) ---------- ---------- 1670371 357834 1670372 370487 1670373 403604 1670374 404679 在数据处理的时候。可以看到dbms_parallel_execute后台启用的处理进程和并行还是有一些不同的。 启用了4个并行之后，看到都是j00这样的进程。

top - 06:31:03 up 1 day, 5:21, 2 users, load average: 3.97, 1.55, 0.61 Tasks: 167 total, 4 running, 163 sleeping, 0 stopped, 0 zombie Cpu(s): 60.7%us, 7.7%sy, 0.0%ni, 1.0%id, 28.9%wa, 0.2%hi, 1.5%si, 0.0%st Mem: 2030124k total, 1293220k used, 736904k free, 358400k buffers Swap: 4063224k total, 0k used, 4063224k free, 476552k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 32630 ora11g 20 0 530m 67m 64m D 55.2 3.4 0:48.34 ora_j000_TEST01 32634 ora11g 20 0 529m 60m 57m R 38.4 3.0 0:46.88 ora_j002_TEST01 32632 ora11g 20 0 529m 62m 59m R 24.8 3.2 0:44.54 ora_j001_TEST01 32636 ora11g 20 0 529m 59m 56m R 17.5 3.0 0:44.88 ora_j003_TEST01 2295 ora11g 20 0 541m 79m 67m D 1.3 4.0 0:11.93 ora_dbw0_TEST01 32706 ora11g 20 0 14940 1240 904 R 1.0 0.1 0:00.39 top -c 825 root 20 0 0 0 0 S 0.3 0.0 0:18.85 [jbd2/sdb3-8] 如果调高parallel从4到16，可以看到j00的进程相应的增加了。

top - 06:32:59 up 1 day, 5:23, 2 users, load average: 1.31, 1.29, 0.63 Tasks: 182 total, 4 running, 178 sleeping, 0 stopped, 0 zombie Cpu(s): 77.2%us, 21.2%sy, 0.0%ni, 1.2%id, 0.2%wa, 0.0%hi, 0.2%si, 0.0%st Mem: 2030124k total, 1345284k used, 684840k free, 358500k buffers Swap: 4063224k total, 0k used, 4063224k free, 476800k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 404 ora11g 20 0 530m 39m 36m R 23.4 2.0 0:01.75 ora_j010_TEST01 392 ora11g 20 0 529m 33m 30m S 17.1 1.7 0:01.39 ora_j004_TEST01 402 ora11g 20 0 529m 31m 28m R 16.1 1.6 0:00.86 ora_j009_TEST01 384 ora11g 20 0 530m 35m 32m S 12.2 1.8 0:01.21 ora_j000_TEST01 386 ora11g 20 0 529m 32m 29m S 11.9 1.6 0:01.11 ora_j001_TEST01 422 root 20 0 122m 19m 10m S 11.5 1.0 0:00.35 /u04/app/11.2.0/grid/bin/crsctl.bin check has 394 ora11g 20 0 529m 31m 29m S 10.5 1.6 0:00.86 ora_j005_TEST01 410 ora11g 20 0 530m 31m 28m R 10.5 1.6 0:00.63 ora_j013_TEST01 408 ora11g 20 0 529m 32m 29m S 9.6 1.6 0:00.93 ora_j012_TEST01 388 ora11g 20 0 529m 32m 29m S 8.9 1.6 0:01.14 ora_j002_TEST01 398 ora11g 20 0 530m 32m 29m S 8.6 1.6 0:00.98 ora_j007_TEST01 390 ora11g 20 0 529m 31m 28m S 7.6 1.6 0:00.74 ora_j003_TEST01 396 ora11g 20 0 529m 32m 29m S 7.2 1.6 0:01.04 ora_j006_TEST01 406 ora11g 20 0 530m 30m 27m S 5.6 1.5 0:00.49 ora_j011_TEST01 414 ora11g 20 0 529m 29m 26m S 5.6 1.5 0:00.45 ora_j015_TEST01 400 ora11g 20 0 529m 30m 27m S 4.9 1.5 0:00.64 ora_j008_TEST01 412 ora11g 20 0 530m 30m 27m S 4.6 1.5 0:00.63 ora_j014_TEST01