rebuild online 创建时，会话被Kill修复索引测试

rebuild online 创建时，会话被Kill修复索引

1.0实验目的：日常运维经常create index online,但是期间被kill会导致索引再次创建失败，测试解决该问题

2.0测试流程：
创建一个测试表，创建一个测试Local索引，记录创建时间(有足够Kill session time)
查询会话ID
创建测试分区索引
kill 会话
查询视图，该索引状态
drop 索引尝试
再次创建索引进行测试
使用dbms包进行清理

3.0创建测试表，测试索引创建时间
#drop table scott.t purge;
SCOTT > create table scott.t(id int,name varchar2(20)) partition by hash(id) partitions 4;
set timing on
declare
v_id int;
begin
for v_id in 1 .. 5000000
loop
insert into scott.t values(v_id,'test'||v_id);
end loop;
commit;
end;
/
Elapsed: 00:10:27.90
create index scott.t_p_ind on scott.t(id) local;
Elapsed: 00:00:20.25

drop index scott.t_p_ind;

3.1 kill session script
SYS > select inst_id,sid,serial#,paddr from gv$session where sid =(select sid from v$mystat where rownum=1);
INST_ID SID SERIAL# PADDR
---------- ---------- ---------- ----------------
1 41 2799 0000000073A1C4A8
select '!ps -ef|grep '||pro.spid ||' '|| '!kill -9 '||pro.spid from v$session ses,v$process pro where ses.sid in( 41 ) and ses.paddr=pro.addr

'!PS-EF|GREP'||PRO.SPID||''||'!KILL-9'||PRO.SPID
--------------------------------------------------------------------------------
!ps -ef|grep 3868 !kill -9 3868

3.2 create index SQL execute
create index scott.t_p_ind on scott.t(id) online;

3.3 kill create index session
!kill -9 3868

create index scott.t_p_ind on scott.t(id) online
ERROR at line 1:
ORA-03113: end-of-file on communication channel
Process ID: 3868
Session ID: 41 Serial number: 2799

3.4查询索引被kill后的状态
SYS > select owner,index_name,status from dba_indexes where owner='SCOTT' and index_name='T_P_IND';
OWNER INDEX_NAME STATUS
--------- ---------------- --------
SCOTT T_P_IND VALID

####创建一个其它列索引，使用hint 是否强制走索引？ [failed] 不重要，可忽略
SYS > create index scott.test_ind on scott.t(name);
set autotrace on
select count(name) from scott.t;
COUNT(*)
----------
5000000
-----------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time | Pstart| Pstop |
-----------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 4389 (1)| 00:00:53 | | |
| 1 | RESULT CACHE | cnrwbkdx2qdpa9s91jr2ka3jzb | | | | | |
| 2 | SORT AGGREGATE | | 1 | | | | |
| 3 | PARTITION HASH ALL| | 4759K| 4389 (1)| 00:00:53 | 1 | 4 |
| 4 | TABLE ACCESS FULL| T | 4759K| 4389 (1)| 00:00:53 | 1 | 4 |
-----------------------------------------------------------------------------------------------------------

select /*+ INDEX(TEST_IND) */ count(*) from scott.t;
-----------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time | Pstart| Pstop |
-----------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 4389 (1)| 00:00:53 | | |
| 1 | RESULT CACHE | cnrwbkdx2qdpa9s91jr2ka3jzb | | | | | |
| 2 | SORT AGGREGATE | | 1 | | | | |
| 3 | PARTITION HASH ALL| | 4759K| 4389 (1)| 00:00:53 | 1 | 4 |
| 4 | TABLE ACCESS FULL| T | 4759K| 4389 (1)| 00:00:53 | 1 | 4 |
-----------------------------------------------------------------------------------------------------------

SYS > desc scott.t
Name Type
--------------------------------------------
ID NUMBER(38)
NAME VARCHAR2(20)

alter table scott.t modify name not null;
再次重复上述操作，还是全表扫描
收集统计信息
analyze table SCOTT.T compute statistics
*
ERROR at line 1:
ORA-08104: this index object 91965 is being online built or rebuilt
。。。。。。出现之前online创建索引报错的信息，无法收集该索引统计信息

单独收集索引信息
Analyze index scott.TEST_IND estimate statistics
Index analyzed.

再次重复上述查询
count(*)还是未走索引，11g Oracle优化器使用的是基于成本的考量，因此我们无法强制改变，hint只是一定范围内的提示，而并非100%指定

修改上述思路，查询一条记录
SYS > select count(*) from scott.t t where name LIKE 'test999999';

COUNT(*)
----------
1
------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 12 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 12 | | |
|* 2 | INDEX RANGE SCAN| TEST_IND | 1 | 12 | 3 (0)| 00:00:01 |
------------------------------------------------------------------------------

drop index scott.TEST_IND;
SYS > select count(*) from scott.t t where name LIKE 'test999999';

COUNT(*)
----------
1
-------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
-------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 12 | 4398 (1)| 00:00:53 | | |
| 1 | RESULT CACHE | 0n8qmskbrjq8sbj8x2d00ugtz4 | | | | | | |
| 2 | SORT AGGREGATE | | 1 | 12 | | | | |
| 3 | PARTITION HASH ALL| | 186 | 2232 | 4398 (1)| 00:00:53 | 1 | 4 |
|* 4 | TABLE ACCESS FULL| T | 186 | 2232 | 4398 (1)| 00:00:53 | 1 | 4 |
-------------------------------------------------------------------------------------------------------------------
#####

回归本次实验目的，测试的是online create index kill 之后，对索引问题的修复
本次索引虽然状态是好的，但实际索引是有问题的
SYS > select owner,index_name,status from dba_indexes where owner='SCOTT' and index_name='T_P_IND';
OWNER INDEX_NAME STATUS
--------- ---------------- --------
SCOTT T_P_IND VALID

SYS > select count(*) from scott.t where id=1;
COUNT(*)
----------
1
----------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
----------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 13 | 1389 (22)| 00:00:17 | | |
| 1 | RESULT CACHE | d74bgy2vpgmj18rsc3c4w4dygh | | | | | | |
| 2 | SORT AGGREGATE | | 1 | 13 | | | | |
| 3 | PARTITION HASH SINGLE| | 44 | 572 | 1389 (22)| 00:00:17 | 4 | 4 |
|* 4 | TABLE ACCESS FULL | T | 44 | 572 | 1389 (22)| 00:00:17 | 4 | 4 |
----------------------------------------------------------------------------------------------------------------------

如果索引状态正常，执行计划将走索引，而非table access full 全表扫描，如何解决索引问题？

3.5drop 索引尝试
drop index soctt.T_P_IND;
drop index soctt.T_P_IND
*
ERROR at line 1:
ORA-01418: specified index does not exist

再次创建索引进行测试

SYS > create index scott.t_t on scott.t(id);
create index scott.t_t on scott.t(id)
*
ERROR at line 1:
ORA-01408: such column list already indexed

对该索引rebuild重建

alter index scott.t_p_ind rebuild online;
ERROR at line 1:
ORA-08104: this index object 91970 is being online built or rebuilt

查询报错解释说明，该索引正在重建
SYS > !oerr ora 08104
08104, 00000, "this index object %s is being online built or rebuilt"
// *Cause: the index is being created or rebuild or waited for recovering
// from the online (re)build
// *Action: wait the online index build or recovery to complete

MOS （ID 375856.1)
在运行在线索引重建时，会话被终止或以其他方式异常终止。您现在正在尝试再次运行索引重建并抛出错误：
已创建dbms_repair.online_index_clean函数以清除联机索引重建。使用dbms_repair.online_index_clean函数来解决此问题。

参考LOB
https://blog.csdn.net/haibusuanyun/article/details/50236057

create /rebuild index online时会修改数据库字典表obj$,并在该索引用户下创建表sys_journal_obj#和ind$、ind_online$表里
如果服务器进程在语句执行过程中意外终止的话，可能会导致相关在ind$标记位信息及在线日志中间表不能及时处理及清除（清除动作一般有smon进程处理,如果重建过程异常中断,smon会清理重建痕迹,但是如果系统非常繁忙导致smon无法清除,或dml操作频繁,导致smon无法获取相关表上的锁,从而造成无法清理痕迹,当再次重建索引或对表进行dml操作会报本篇提示错误），这将导致对该索引的后续操作因ora-08104错误而无法继续，如果是分区表，索引是global，在添加分区也无法继续。

4.0解决问题
4.1表没有tm锁，调用存储过程进行清理
4.2表存在事务，需等待事务结束
4.3存在大量的rebuild online被结束，批量操作

4.1表没有tm锁，调用存储过程进行清理
--查询字典表中的记录
select i.obj#, i.flags, u.name, o.name, o.type#
from sys.obj$ o, sys.user$ u, sys.ind_online$ i
where (bitand(i.flags, 256) = 256 or bitand(i.flags, 512) = 512)
and (not ((i.type# = 9) and bitand(i.flags, 8) = 8))
and o.obj# = i.obj#
and o.owner# = u.user#;
OBJ# FLAGS NAME NAME TYPE#
---------- ---------- ---------------- -------------------
91970 256 SCOTT T_P_IND 1

--调用dbms_repair.online_index_clean
declare
a boolean;
begin
a := dbms_repair.online_index_clean(91970,dbms_repair.lock_wait);
end;
/

SYS > select owner,index_name,status from dba_indexes where owner='SCOTT' and index_name='T_T';
no rows selected

4.2表存在事务，需等待事务结束
--如果此时，表存在TM 3 dml事务？
create index scott.t_t on scott.t(id) online;
*
ERROR at line 1:
ORA-03113: end-of-file on communication channel
Process ID: 4245
Session ID: 65 Serial number: 115
--再次查询视图
OBJ# FLAGS NAME NAME TYPE#
---------- ---------- ------------------------------ ------------------------------ ----------
91974 256 SCOTT T_T 1
--dml操作

SYS > delete scott.t where rownum<2;
1 row deleted.
--调用dbms_repair.online_index_clean
declare
a boolean;
begin
a := dbms_repair.online_index_clean(91974,dbms_repair.lock_wait);
end;
/ --会话hang住

Alert log
Sat May 12 10:03:52 2018
online index (re)build cleanup: objn=91974 maxretry=2000 forever=0

--
select object_name,s.sid,s.serial#,s.username,' '||'!kill -9 '||p.spid as"kill"
from v$locked_object l , dba_objects o , v$session s , v$process p
where l.object_id=o.object_id and l.session_id=s.sid and s.paddr=p.addr and o.owner='SCOTT' and o.object_name='T';
OBJECT_NAM SID SERIAL# USERNAME kill
---------- ---------- ---------- ---------- --------------------
T 40 2437 SYS !kill -9 4114
T 40 2437 SYS !kill -9 4114
--kill session
可释放资源

--4.3存在大量的rebuild online被结束，批量操作

declare
isClean boolean;
begin
isClean := FALSE;
while isClean=FALSE loop
isClean := dbms_repair.online_index_clean(dbms_repair.all_index_id,
dbms_repair.lock_wait);
dbms_lock.sleep(2);
end loop;
exception
when others then
RAISE;
end;

--测试创建两个不同用户下的online失效

create index scott.t_t on scott.t(id) online;
kill session
HR > create table hr.t1 as select * from scott.t;
select distinct sid from v$mystat;
create index hr.c1 on hr.t1(id) online;
kill session

--查询视图记录
OBJ# FLAGS NAME NAME TYPE#
---------- ---------- ----------------- --------------------
91977 256 SCOTT T_T 1
91987 256 HR C1 1
--执行存储过程
PL/SQL procedure successfully completed.
--Alert日志记录
Sat May 12 10:19:29 2018
online index (re)build cleanup: objn=91977 maxretry=2000 forever=0
online index (re)build cleanup: objn=91987 maxretry=2000 forever=0

--重建完毕，如果存在锁资源，相信也无法申请，kill或等待即可