row lock waits的两种情况
以下测试基于版本9207。
情形一：并发更新非唯一索引的索引字段引起的row lock waits在表上，不在索引上
SQL> create table test as select object_id,object_name from dba_objects;
Table created.
SQL> create index test_idx on test(object_id);
Index created.
SQL> update test set object_id=12345 where object_id=99998;
1 row updated.
SQL> select statistic_name,value from v$segment_statistics where owner=’TEST’ and object_name=’TEST’ and statistic_name=’row lock waits’;
STATISTIC_NAME VALUE
—————————————————————- ———-
row lock waits 0
SQL> select statistic_name,value from v$segment_statistics where owner=’TEST’ and object_name=’TEST_IDX’ and statistic_name=’row lock waits’;
STATISTIC_NAME VALUE
—————————————————————- ———-
row lock waits 0
另一session;
SQL> update test set object_id=12345 where object_id=99998; hang住，此时除了有enqueue等待外，还有表上的row lock waits 等待，这个等待发生在表上，而不是索引上。
第三个session:
SQL> select statistic_name,value from v$segment_statistics where owner=’TEST’ and object_name=’TEST’;
STATISTIC_NAME VALUE
—————————————————————- ———-
logical reads 61424
buffer busy waits 0
db block changes 29088
physical reads 337
physical writes 337
physical reads direct 0
physical writes direct 337
global cache cr blocks served 0
global cache current blocks served 0
ITL waits 0
row lock waits 1
SQL> select statistic_name,value from v$segment_statistics where owner=’TEST’ and object_name=’TEST_IDX’;
STATISTIC_NAME VALUE
—————————————————————- ———-
logical reads 96
buffer busy waits 0
db block changes 64
physical reads 2
physical writes 129
physical reads direct 0
physical writes direct 129
global cache cr blocks served 0
global cache current blocks served 0
ITL waits 0
row lock waits 0
情形二：索引块分裂引起的row lock waits 等待在索引上，而不在表上
session 1:
SQL> show parameter db_block_size
NAME TYPE VALUE
———————————— ———– ——————————
db_block_size integer 4096
SQL> create table test(x number,y char(20),z date,q varchar2(2000));
Table created.
SQL> create index test_idx on test(q,z);
Index created.
SQL> begin
2 for i in 1..5000
3 loop
4 insert into test values(1162,’1060000abcdefg’, sysdate, rpad(‘x’,1000,’x'));
5 end loop;
6 end;
7 /
PL/SQL procedure successfully completed.
session 2:
与session 并发插入
SQL> begin
2 for i in 1..5000
3 loop
4 insert into test values(1162,’1060000abcdefg’, sysdate, rpad(‘x’,1000,’x'));
5 end loop;
6 end;
7 /
PL/SQL procedure successfully completed.
session 3:
两个session 运行插入之前:
OBJECT_NAME STATISTIC_NAME VALUE
—————————— ———————————————— ———-
TEST_IDX row lock waits 0
TEST row lock waits 0
运行插入过程中：
SQL> /
OBJECT_NAME STATISTIC_NAME VALUE
—————————— ———————————————— ———-
TEST_IDX row lock waits 7
TEST row lock waits 0
SQL> /
OBJECT_NAME STATISTIC_NAME VALUE
—————————— ——————————————– ———-
TEST_IDX row lock waits 20
TEST row lock waits 0
SQL> /
OBJECT_NAME STATISTIC_NAME VALUE
—————————— ——————————————– ———-
TEST_IDX row lock waits 481
TEST row lock waits 0
SQL> /
OBJECT_NAME STATISTIC_NAME VALUE
—————————— ——————————————— ———-
TEST_IDX row lock waits 906
TEST row lock waits 0
可以看到此时的row lock waits等待有多达906个，这个row lock waits统计经常伴随着enq: TX – index contention等待事件的发生。
基于以上两种情况，所以我们有时候看到row lock waits在表上，有时候在索引上，所以要具体原因具体分析，然后才能对症解决。
不错，以实际演示问题！
原文地址： http://www.dbasky.com/oracle/update_index_row_lock_waits.html