Mapping Files to Physical Devices

Components of File Mapping

The following figure shows the components of the file mapping mechanism.

Figure 13-1 Components of File Mapping

Description of "Figure 13-1 Components of File Mapping"

The following sections briefly describes these components and how they work together to populate the mapping views:

• FMON

• External Process (FMPUTL)

• Mapping Libraries

Mapping Structures

The mapping structures and the Oracle Database representation of these structures are described in this section. You will need to understand this information in order to interpret the information in the mapping views.

The following are the primary structures that compose the mapping information:

• Files

  A file mapping structure provides a set of attributes for a file, including file size, number of file system extents that the file is composed of, and the file type.

• File system extents

  A file system extent mapping structure describes a contiguous chunk of blocks residing on one element. This includes the device offset, the extent size, the file offset, the type (data or parity), and the name of the element where the extent resides.

  Note:

  File system extents are not the same as Oracle Database extents. File system extents are physical contiguous blocks of data written to a device as managed by the file system. Oracle Database extents are logical structures managed by the database, such as tablespace extents.

• Elements

  An element mapping structure is the abstract mapping structure that describes a storage component within the I/O stack. Elements may be mirrors, stripes, partitions, RAID5, concatenated elements, and disks. These structures are the mapping building blocks.

• Subelements

  A subelement mapping structure describes the link between an element and the next elements in the I/O mapping stack. This structure contains the subelement number, size, the element name where the subelement exists, and the element offset.

All of these mapping structures are illustrated in the following example.

Example of Mapping Structures

Consider an Oracle Database which is composed of two data files X and Y. Both files X and Y reside on a file system mounted on volume A. File X is composed of two extents while file Y is composed of only one extent.

The two extents of File X and the one extent of File Y both map to Element A. Element A is striped to Elements B and C. Element A maps to Elements B and C by way of Subelements B0 and C1, respectively.

Element B is a partition of Element D (a physical disk), and is mapped to Element D by way of subelement D0.

Element C is mirrored over Elements E and F (both physical disks), and is mirrored to those physical disks by way of Subelements E0 and F1, respectively.

All of the mapping structures are illustrated in Figure 13-2.

Figure 13-2 Illustration of Mapping Structures

Description of "Figure 13-2 Illustration of Mapping Structures"

Note that the mapping structures represented are sufficient to describe the entire mapping information for the Oracle Database instance and consequently to map every logical block within the file into a (element name, element offset) tuple (or more in case of mirroring) at each level within the I/O stack.

Configuration ID

The configuration ID captures the version information associated with elements or files. The vendor library provides the configuration ID and updates it whenever a change occurs. Without a configuration ID, there is no way for the database to tell whether the mapping has changed.

There are two kinds of configuration IDs:

• Persistent

  These configuration IDs are persistent across instance shutdown

• Non-persistent

  The configuration IDs are not persistent across instance shutdown. The database is only capable of refreshing the mapping information while the instance is up.

Enabling File Mapping

The following steps enable the file mapping feature:

1. Ensure that a valid filemap.ora file exists in the /opt/ORCLfmap/prot1_32/etc directory for 32-bit platforms, or in the /opt/ORCLfmap/prot1_64/etc directory for 64-bit platforms.

   Caution:

   While the format and content of the filemap.ora file is discussed here, it is for informational reasons only. The filemap.ora file is created by the database when your system is installed. Until such time that vendors supply there own libraries, there will be only one entry in the filemap.ora file, and that is the Oracle-supplied EMC library. This file should be modified manually by uncommenting this entry only if an EMC Symmetrix array is available.

   The filemap.ora file is the configuration file that describes all of the available mapping libraries. FMON requires that a filemap.ora file exists and that it points to a valid path to mapping libraries. Otherwise, it will not start successfully.

   The following row needs to be included in filemap.ora for each library:

   lib=vendor_name:mapping_library_path

   where:

   • vendor_name should be Oracle for the EMC Symmetric library

   • mapping_library_path is the full path of the mapping library

   Note that the ordering of the libraries in this file is extremely important. The libraries are queried based on their order in the configuration file.

   The file mapping service can be even started even if no mapping libraries are available. The filemap.ora file still needs to be present even though it is empty. In this case, the mapping service is constrained in the sense that new mapping information cannot be discovered. Only restore and drop operations are allowed in such a configuration.

2. Set the FILE_MAPPING initialization parameter to TRUE.

   The instance does not have to be shut down to set this parameter. You can set it using the following ALTER SYSTEM statement:

   ALTER SYSTEM SET FILE_MAPPING=TRUE;
   

3. Invoke the appropriate DBMS_STORAGE_MAP mapping procedure. You have two options:

   • In a cold startup scenario, the Oracle Database is just started and no mapping operation has been invoked yet. You execute the DBMS_STORAGE_MAP.MAP_ALL procedure to build the mapping information for the entire I/O subsystem associated with the database.

   • In a warm start scenario where the mapping information is already built, you have the option to invoke the DBMS_STORAGE_MAP.MAP_SAVE procedure to save the mapping information in the data dictionary. (Note that this procedure is invoked in DBMS_STORAGE_MAP.MAP_ALL() by default.) This forces all of the mapping information in the SGA to be flushed to disk.

     Once you restart the database, use DBMS_STORAGE_MAP.RESTORE() to restore the mapping information into the SGA. If needed, DBMS_STORAGE_MAP.MAP_ALL() can be called to refresh the mapping information.

Using the DBMS_STORAGE_MAP Package

The DBMS_STORAGE_MAP package enables you to control the mapping operations. The various procedures available to you are described in the following table.

Obtaining Information from the File Mapping Views

Mapping information generated by DBMS_STORAGE_MAP package is captured in dynamic performance views. Brief descriptions of these views are presented here.

However, the information generated by the DBMS_STORAGE_MAP.MAP_OBJECT procedure is captured in a global temporary table named MAP_OBJECT. This table displays the hierarchical arrangement of storage containers for objects. Each row in the table represents a level in the hierarchy. A description of the MAP_OBJECT table follows.

Example 1: Map All Database Files that Span a Device

The following query returns all Oracle Database files associated with the /dev/vx/rdmp/c2t1d1s2 host device:

SELECT UNIQUE me.ELEM_NAME, mf.FILE_NAME
   FROM V$MAP_FILE_IO_STACK fs, V$MAP_FILE mf, V$MAP_ELEMENT me
   WHERE mf.FILE_MAP_IDX = fs.FILE_MAP_IDX
   AND me.ELEM_IDX = fs.ELEM_IDX
   AND me.ELEM_NAME = '/dev/vx/rdmp/c2t1d1s2';
 

The query results are:

ELEM_NAME                 FILE_NAME
------------------------  --------------------------------
/dev/vx/rdmp/c2t1d1s2     /oracle/dbs/t_db1.f
/dev/vx/rdmp/c2t1d1s2     /oracle/dbs/t_db2.f

Example 2: Map a File into Its Corresponding Devices

The following query displays a topological graph of the /oracle/dbs/t_db1.f datafile:

WITH fv AS 
  (SELECT FILE_MAP_IDX, FILE_NAME FROM V$MAP_FILE 
   WHERE FILE_NAME = '/oracle/dbs/t_db1.f')
SELECT fv.FILE_NAME, LPAD(' ', 4 * (LEVEL - 1)) || el.ELEM_NAME ELEM_NAME 
   FROM V$MAP_SUBELEMENT sb, V$MAP_ELEMENT el, fv, 
     (SELECT UNIQUE ELEM_IDX FROM V$MAP_FILE_IO_STACK io, fv 
      WHERE io.FILE_MAP_IDX = fv.FILE_MAP_IDX) fs
   WHERE el.ELEM_IDX = sb.CHILD_IDX
   AND fs.ELEM_IDX = el.ELEM_IDX
   START WITH sb.PARENT_IDX IN
     (SELECT DISTINCT ELEM_IDX 
      FROM V$MAP_FILE_EXTENT fe, fv
      WHERE fv.FILE_MAP_IDX = fe.FILE_MAP_IDX)
   CONNECT BY PRIOR sb.CHILD_IDX = sb.PARENT_IDX;

The resulting topological graph is:

FILE_NAME                    ELEM_NAME
-----------------------      -------------------------------------------------
/oracle/dbs/t_db1.f          _sym_plex_/dev/vx/rdsk/ipfdg/ipf-vol1_-1_-1
/oracle/dbs/t_db1.f              _sym_subdisk_/dev/vx/rdsk/ipfdg/ipf-vol1_0_0_0
/oracle/dbs/t_db1.f                  /dev/vx/rdmp/c2t1d0s2
/oracle/dbs/t_db1.f                      _sym_symdev_000183600407_00C
/oracle/dbs/t_db1.f                          _sym_hyper_000183600407_00C_0
/oracle/dbs/t_db1.f                          _sym_hyper_000183600407_00C_1
/oracle/dbs/t_db1.f              _sym_subdisk_/dev/vx/rdsk/ipfdg/ipf-vol1_0_1_0
/oracle/dbs/t_db1.f                  /dev/vx/rdmp/c2t1d1s2
/oracle/dbs/t_db1.f                      _sym_symdev_000183600407_00D
/oracle/dbs/t_db1.f                          _sym_hyper_000183600407_00D_0
/oracle/dbs/t_db1.f                          _sym_hyper_000183600407_00D_1

Example 3: Map a Database Object

This example displays the block distribution at all levels within the I/O stack for the scott.bonus table.

A MAP_OBJECT() operation must first be executed as follows:

EXECUTE DBMS_STORAGE_MAP.MAP_OBJECT('BONUS','SCOTT','TABLE');

The query is as follows:

SELECT io.OBJECT_NAME o_name, io.OBJECT_OWNER o_owner, io.OBJECT_TYPE o_type,
       mf.FILE_NAME, me.ELEM_NAME, io.DEPTH,
      (SUM(io.CU_SIZE * (io.NUM_CU - DECODE(io.PARITY_PERIOD, 0, 0, 
                         TRUNC(io.NUM_CU / io.PARITY_PERIOD)))) / 2) o_size
   FROM MAP_OBJECT io, V$MAP_ELEMENT me, V$MAP_FILE mf
   WHERE io.OBJECT_NAME =  'BONUS'
   AND   io.OBJECT_OWNER = 'SCOTT'
   AND   io.OBJECT_TYPE =  'TABLE'
   AND   me.ELEM_IDX = io.ELEM_IDX
   AND   mf.FILE_MAP_IDX = io.FILE_MAP_IDX
   GROUP BY io.ELEM_IDX, io.FILE_MAP_IDX, me.ELEM_NAME, mf.FILE_NAME, io.DEPTH,
         io.OBJECT_NAME, io.OBJECT_OWNER, io.OBJECT_TYPE
   ORDER BY io.DEPTH;

The following is the result of the query. Note that the o_size column is expressed in KB.

O_NAME O_OWNER O_TYPE  FILE_NAME            ELEM_NAME                      DEPTH   O_SIZE
------ ------- ------  -------------------  -----------------------------  ------  ------
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  /dev/vx/dsk/ipfdg/ipf-vol1          0      20
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_plex_/dev/vx/rdsk/ipf          1      20
                                            pdg/if-vol1_-1_-1
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_subdisk_/dev/vx/rdsk/          2      12
                                            ipfdg/ipf-vol1_0_1_0
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_subdisk_/dev/vx/rdsk/ipf       2       8
                                            dg/ipf-vol1_0_2_0 
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  /dev/vx/rdmp/c2t1d1s2               3      12
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  /dev/vx/rdmp/c2t1d2s2               3       8
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_symdev_000183600407_00D        4      12
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_symdev_000183600407_00E        4       8
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_hyper_000183600407_00D_0       5      12
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_hyper_000183600407_00D_1       5      12
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_hyper_000183600407_00E_0       6       8
BONUS  SCOTT   TABLE   /oracle/dbs/t_db1.f  _sym_hyper_000183600407_00E_1       6       8