HP NonStop RDF, Руководство по управлению

Страница: 163 / 478

a backup index to be different than that of the primary index. In such a case, the index rows
transmitted from the primary system to the backup system will be corrupt with regard to their
key values. Although the records are physically present in the index on the backup system,
NonStop SQL/MP does not see them because the actual key specifier value does not match the
expected one. Consequently, a FUP INFO
index
, STAT display will show the correct number
of records for the index, but a SQLCI SELECT COUNT (*) FROM
index
command will return
fewer rows for the index than indicated by the FUP INFO command. The row count continues
to grow in the base table, but remains the same for the index.
You can avoid this problem by always using the KEYTAG clause in the CREATE INDEX command
to define a meaningful key specifier for each index you create.
If you encounter the problem described above, use SQLCI to DROP and re-CREATE the offending
indexes, doing so in the proper creation order. The following annotated output illustrates the
necessary index creation order:
>SQLCI FILEINFO $DATA.RDFSQL.MASTER, DETAIL
$DATA.RDFSQL.MASTER
SQL BASE TABLE
CATALOG $DATA.RDFSQL
VERSION 2
TYPE K
EXT ( 16 PAGES, 64 PAGES, MAXEXTENTS 160 )
REC 416
PACKED REC 415
BLOCK 4096
KEY ( COLUMN 0, OFFSET 0, LENGTH 4, ASC )
INDEX ( 1, $DATA.RDFSQL.MASTXYZ, <<create this index first
COLUMN 18, OFFSET 54, LENGTH 2, ASC.
COLUMN 19, OFFSET 56, LENGTH 2, ASC.
NOT UNIQUE )
INDEX ( 2, $DATA.RDFSQL.MASTABC, <<create this index second
COLUMN 88, OFFSET 300, LENGTH 15, ASC.
COLUMN 87, OFFSET 285, LENGTH 15, ASC.
NOT UNIQUE )
AUDIT
BUFFERED
AUDITCOMPRESS
SECURITY (RWEP); NCNC
MODIF: 27 Dec 1997, 20:01
CREATION DATE: 02 Dec 1997, 12:37
REDEFINITION DATE: 10 Jan 1998, 14:46
LAST OPEN: 10 Jan 1998, 14:46
EOF 466944 (2.2% USED)
EXTENTS ALLOCATED: 160
INDEX LEVELS: 1
PARTITION ARRAY STANDARD
Partition Key Changes
If you change a key for any partition on the primary system, you must also change the key for
the corresponding partition on the backup system.
Table Purges
If you use the SQLCI PURGE command to purge a protected table from the primary system, you
must also purge the corresponding table from the backup system. You should not purge a table
on the backup system until you are sure RDF has completed all processing on the table. The
processing is complete when the RTD time is zero for the updater process associated with the
table’s volume. To check the RTD time, issue a STATUS RDF command.
Making Changes to Database Structures 163