问题描述：
经常有人问，我分区表里的索引到底是分区索引还是非分区索引？
因为是否是分区索引涉及到detach分区的时候是否会耗费大量的时间做异步索引清理：如果是非分区索引，则异步索引清理需要大量时间。
总体结论：
--对于唯一索引或者主健，如果包含了分区健，则默认是分区索引；如果不包含分区健，则默认是非分区索引。
--对于非唯一索引，默认都是分区索引。
测试过程
DB2版本为10.5
$ db2 "create table p1 (col1 int not null, col2 int not null, col3 int not null) partition by range(col2)(partition part1 starting 1 ending 5, partition part2 starting 6 ending 10, partition part3 starting 11 ending 15)"
1. 唯一索引
$ db2 "create unique index u_idx1 on p1 (col1)"
$ db2 "create unique index u_idx1_2 on p1 (col1,col2)"
$ db2 "create unique index u_idx1_3 on p1 (col1,col3)"
$ db2look -d sample -a -e -t p1
CREATE UNIQUE INDEX "DB2TST "."U_IDX1" ON "DB2TST "."P1"
("COL1" ASC)
NOT PARTITIONED IN "TBS2"
COMPRESS NO
INCLUDE NULL KEYS ALLOW REVERSE SCANS;
-- DDL Statements for Indexes on Table "DB2TST "."P1"
SET SYSIBM.NLS_STRING_UNITS = 'SYSTEM';
CREATE UNIQUE INDEX "DB2TST "."U_IDX1_2" ON "DB2TST "."P1"
("COL1" ASC,
"COL2" ASC)
PARTITIONED
COMPRESS NO
INCLUDE NULL KEYS ALLOW REVERSE SCANS;
-- DDL Statements for Indexes on Table "DB2TST "."P1"
SET SYSIBM.NLS_STRING_UNITS = 'SYSTEM';
CREATE UNIQUE INDEX "DB2TST "."U_IDX1_3" ON "DB2TST "."P1"
("COL1" ASC,
"COL3" ASC)
NOT PARTITIONED IN "TBS2"
COMPRESS NO
INCLUDE NULL KEYS ALLOW REVERSE SCANS;
结论：对于唯一索引，如果索引中包含了分区健，默认是分区索引。如果索引中不包含分区健，默认是非分区索引。
问题：如果是不包含分区健的唯一索引，想做成分区的怎么破？答：没有办法，创建的时候会报错 SQL20303N
$ db2 "create unique index u_idx3 on p1 (col3) partitioned"
DB21034E The command was processed as an SQL statement because it was not a
valid Command Line Processor command. During SQL processing it returned:
SQL20303N The partitioned unique index was not created because either the
index definition did not include all of the partitioning columns, or the index
was being created over XML data. SQLSTATE=42990
2. 主健
删表重建
$ db2 "alter table p1 add primary key(col1)"
$ db2 "select varchar(INDNAME,40) as idxname from syscat.indexpartitions where TABNAME='P1'"
IDXNAME
----------------------------------------
0 record(s) selected.
$ db2 "alter table p1 drop primary key"
$ db2 "alter table p1 add primary key(col1,col2)"
$ db2 "select varchar(INDNAME,40) as idxname from syscat.indexpartitions where TABNAME='P1'"
IDXNAME
----------------------------------------
SQL200122112144130
SQL200122112144130
SQL200122112144130
3 record(s) selected.
结论：主健实际上是唯一索引，因此效果跟唯一索引类似，即：
如果主健中包含了分区健，则主健对应索引是分区索引。 如果主健中不包含分区健，则索引是非分区索引。
3. 非唯一索引
删表重建
$ db2 "create index idx1 on p1 (col1)"
$ db2 "create index idx1_2 on p1 (col1,col2)"
$ db2 "create index idx1_3 on p1 (col1,col3)"
$ db2look -d sample -a -e -t p1
CREATE INDEX "DB2TST "."IDX1" ON "DB2TST "."P1"
("COL1" ASC)
PARTITIONED
COMPRESS NO
INCLUDE NULL KEYS ALLOW REVERSE SCANS;
-- DDL Statements for Indexes on Table "DB2TST "."P1"
SET SYSIBM.NLS_STRING_UNITS = 'SYSTEM';
CREATE INDEX "DB2TST "."IDX1_2" ON "DB2TST "."P1"
("COL1" ASC,
"COL2" ASC)
PARTITIONED
COMPRESS NO
INCLUDE NULL KEYS ALLOW REVERSE SCANS;
-- DDL Statements for Indexes on Table "DB2TST "."P1"
SET SYSIBM.NLS_STRING_UNITS = 'SYSTEM';
CREATE INDEX "DB2TST "."IDX1_3" ON "DB2TST "."P1"
("COL1" ASC,
"COL3" ASC)
PARTITIONED
COMPRESS NO
INCLUDE NULL KEYS ALLOW REVERSE SCANS;
结论：非唯一索引默认都是分区索引。