ASM的FAILGROUP

Normal 0 7.8 磅 0 2 false false false EN-US ZH-CN X-NONE 以前一直以为NORMAL REDUNDANCY的话，FAILGROUP就是2，HIGH REDUNDANCY的话，FAILGROUP就是3

其实这是错误的，FAILGROUP是用于将磁盘分组，以保证丢失任何一组FAILGROUP磁盘，数据还是完整的

多用于多阵列，通过ASM来完成冗余的环境。（EXTERNAL只能有一个FAILGROUP）

 

假设有4个阵列，每个阵列3块盘，使用NORMAL REDUNDANCY创建DISKGROUP，4个FAILGROUP

CREATE DISKGROUP dgroup1 NORMAL REDUNDANCY  FAILGROUP controller1 DISK  'ORCL:ASM01', 'ORCL:ASM02', 'ORCL:ASM03'  FAILGROUP controller2 DISK  'ORCL:ASM04', 'ORCL:ASM05', 'ORCL:ASM06'  FAILGROUP controller3 DISK  'ORCL:ASM07', 'ORCL:ASM08', 'ORCL:ASM09'  FAILGROUP controller4 DISK  'ORCL:ASM10', 'ORCL:ASM11', 'ORCL:ASM12';

 

然后在上面创建一个1G的数据文件

CREATE TABLESPACE asmtest DATAFILE '+DGROUP1' SIZE 1024M

 

然后看看数据的冗余程度

SQL> SELECT COUNT(DISTINCT PXN_KFFXP),COUNT(DISTINCT XNUM_KFFXP)   2    FROM X$KFFXP   3   WHERE GROUP_KFFXP = 3   4      AND NUMBER_KFFXP =256   5      AND xnum_kffxp <> 2147483648;   COUNT(DISTINCTPXN_KFFXP) COUNT(DISTINCTXNUM_KFFXP) ------------------------ -------------------------                     2050                      1025 --数据文件为1024M，加上一个BLOCK0，一共1025M，2个冗余   SQL> SELECT COUNT(DISTINCT PXN_KFFXP),COUNT(DISTINCT XNUM_KFFXP)   2    FROM X$KFFXP   3   WHERE GROUP_KFFXP = 3   4      AND NUMBER_KFFXP =256   5      AND xnum_kffxp = 2147483648;   COUNT(DISTINCTPXN_KFFXP) COUNT(DISTINCTXNUM_KFFXP) ------------------------ -------------------------                        3                         1 --记录块分配的元数据冗余3份，其他元数据也是冗余3份

 

假设出现某个failgroup损坏

SQL> SELECT COUNT (DISTINCT xnum_kffxp)   2    FROM X$KFFXP a,v$asm_disk b   3   WHERE GROUP_KFFXP = 3   4     AND b.group_number=3   5     AND NUMBER_KFFXP =256   6     AND a.disk_kffxp = disk_number   7     AND b.failgroup <> 'CONTROLLER&i'   8  ; Enter value for i: 1 old   7:    AND b.failgroup <> 'CONTROLLER&i' new   7:    AND b.failgroup <> 'CONTROLLER1'   COUNT(DISTINCTXNUM_KFFXP) -------------------------                      1026   SQL> / Enter value for i: 2 old   7:    AND b.failgroup <> 'CONTROLLER&i' new   7:    AND b.failgroup <> 'CONTROLLER2'   COUNT(DISTINCTXNUM_KFFXP) -------------------------                      1026   SQL> / Enter value for i: 3 old   7:    AND b.failgroup <> 'CONTROLLER&i' new   7:    AND b.failgroup <> 'CONTROLLER3'   COUNT(DISTINCTXNUM_KFFXP) -------------------------                      1026   SQL> / Enter value for i: 4 old   7:    AND b.failgroup <> 'CONTROLLER&i' new   7:    AND b.failgroup <> 'CONTROLLER4'   COUNT(DISTINCTXNUM_KFFXP) -------------------------                      1026

 

其相关的1026个AU都是可访问的

 

在看看元数据，元数据每个冗余3份，任何一个FAILGROUP损坏，都能保证数据是完整的

SQL> SELECT NUMBER_KFFXP, XNUM_KFFXP, COUNT(*)   2    FROM X$KFFXP   3   WHERE GROUP_KFFXP = 3   4     AND NUMBER_KFFXP < 100   5   GROUP BY NUMBER_KFFXP, XNUM_KFFXP   6   ORDER BY 1, 2;   NUMBER_KFFXP XNUM_KFFXP   COUNT(*) ------------ ---------- ----------            1          0          3            1          1          3            2          0          3            3          0          3            3          1          3            3          2          3            3          3          3            3          4          3            3          5          3            3          6          3            3          7          3            3          8          3            3          9          3            3         10          3            3         11          3            3         12          3            3         13          3            3         14          3            3         15          3            3         16          3            3         17          3            3         18          3            3         19          3            3         20          3            3         21          3            3         22          3            3         23          3            3         24          3            3         25          3            3         26          3            3         27          3            3         28          3            3         29          3            3         30          3            3         31          3            3         32          3            3         33          3            3         34          3            3         35          3            3         36          3            3         37          3            3         38          3            3         39          3            3         40          3            3         41          3            3 2147483648          3            4          0          3            4          1          3            5          0          3            6          0          3   50 rows selected.   SQL> SELECT COUNT(*)   2    FROM (SELECT DISTINCT NUMBER_KFFXP, XNUM_KFFXP   3    FROM X$KFFXP A, V$ASM_DISK B   4   WHERE GROUP_KFFXP = 3   5     AND B.GROUP_NUMBER = 3   6     AND NUMBER_KFFXP < 100   7     AND A.DISK_KFFXP = DISK_NUMBER   8     AND B.FAILGROUP <> 'CONTROLLER&i') ; Enter value for i: 1 old   8:    AND B.FAILGROUP <> 'CONTROLLER&i') new   8:    AND B.FAILGROUP <> 'CONTROLLER1')     COUNT(*) ----------         50   SQL> / Enter value for i: 2 old   8:    AND B.FAILGROUP <> 'CONTROLLER&i') new   8:    AND B.FAILGROUP <> 'CONTROLLER2')     COUNT(*) ----------         50   SQL> / Enter value for i: 3 old   8:    AND B.FAILGROUP <> 'CONTROLLER&i') new   8:    AND B.FAILGROUP <> 'CONTROLLER3')     COUNT(*) ----------         50   SQL> / Enter value for i: 4 old   8:    AND B.FAILGROUP <> 'CONTROLLER&i') new   8:    AND B.FAILGROUP <> 'CONTROLLER4')     COUNT(*) ----------         50

 

注意，默认创建的DISKGROUP，每个磁盘就是一个单独的FAILGROUP

如上面的例子，4个阵列，在没有指定failgroup的情况下，只要一组阵列出现了宕机，数据就可能无法访问，如果指定了failgroup，一组阵列损坏后，数据仍然能保证完整

 

在来看看冗余报告:

SQL> SELECT NAME,   2         TYPE,   3         TOTAL_MB,   4         FREE_MB,   5         REQUIRED_MIRROR_FREE_MB,   6         USABLE_FILE_MB   7    FROM V$ASM_DISKGROUP   8   WHERE GROUP_NUMBER = 3;   NAME       TYPE           TOTAL_MB    FREE_MB REQUIRED_MIRROR_FREE_MB USABLE_FILE_MB ---------- ------------ ---------- ---------- ----------------------- -------------- DGROUP1    NORMAL            12228      12054                    3057           4498

 

如果failgroup失败，那么需要3057M的空间来恢复，大概3块盘

 

如果是不指定failgroup创建DISKGROUP了？

CREATE DISKGROUP dgroup1 NORMAL REDUNDANCY DISK  'ORCL:ASM01', 'ORCL:ASM02', 'ORCL:ASM03', 'ORCL:ASM04', 'ORCL:ASM05', 'ORCL:ASM06',  'ORCL:ASM07', 'ORCL:ASM08', 'ORCL:ASM09', 'ORCL:ASM10', 'ORCL:ASM11', 'ORCL:ASM12';

SQL> /   NAME       TYPE           TOTAL_MB    FREE_MB REQUIRED_MIRROR_FREE_MB USABLE_FILE_MB ---------- ------------ ---------- ---------- ----------------------- -------------- DGROUP1    NORMAL            12228      12057                    1111           5473

 

如果不指定，只允许一个盘损坏，需要1111M来恢复，大概1块盘

 

USABLE_FILE_MB 不代表能使用的磁盘限制，但当该列小于0时，不能正常DROP DISK，无法完成rebalance

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/8242091/viewspace-671590/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/8242091/viewspace-671590/