Oracle ASM Filter Driver (Oracle ASMFD)是位于Oracle ASM磁盘IO路径上的一个内核模块，Oracle ASM使用ASMFD来验证对Oracle ASM磁盘的写IO请求，Oracle ASM Filter Driver会拒绝任何无效的IO请求。此操作消除了意外覆盖Oracle ASM磁盘头的情况。例如，Oracle ASM Filter Driver过滤掉所有可能导致意外覆盖的非Oracle IO。
默认情况下，Oracle ASM Filter Driver (Oracle ASMFD)与Oracle Grid Infrastructure安装一起安装。如果现有的磁盘组未使用ASMFD，则需要将现有的ASM磁盘组中的磁盘进行标识，并且将其迁移到ASMFD (ASM Filter Driver)。
本示例将磁盘组从传统的udev方式切换成AFD方式。
1、以GRID用户运行如下命令，生成ASM磁盘与ASM磁盘组的对应关系。
|
SQL> set lines 200 SQL> col path for a30 SQL> col name for a30 SQL> col header_status for a30 SQL> select name, path,header_status from v$asm_disk order by 1;
NAME PATH HEADER_STATUS ------------------------------ ------------------------------ ------------------ DG_DATA_0000 /dev/oracleasm/disks/mpathj MEMBER DG_DATA_0001 /dev/oracleasm/disks/mpathk MEMBER DG_DATA_0002 /dev/oracleasm/disks/mpathl MEMBER DG_GRID_0000 /dev/oracleasm/disks/mpathh MEMBER DG_GRID_0001 /dev/oracleasm/disks/mpathi MEMBER
SQL> |
2、以GRID用户运行如下命令，获取现有ASM磁盘的搜索路径。
|
[grid@19crac1 ~]$ asmcmd dsget parameter:/dev/oracleasm/disks/* profile:/dev/oracleasm/disks/* [grid@19crac1 ~]$ |
3、以GRID用户运行如下命令，更新ASM磁盘的搜索路径，使Oracle ASMFD能够发现用于迁移的设备。
|
[grid@19crac1 ~]$ asmcmd dsset '/dev/oracleasm/disks/*','AFD:*' [grid@19crac1 ~]$ [grid@19crac1 ~]$ asmcmd dsget parameter:/dev/oracleasm/disks/*, AFD:* profile:/dev/oracleasm/disks/*,AFD:* [grid@19crac1 ~]$ |
4、停止当前节点的CRS集群。
|
[root@19crac1 ~]# crsctl stop crs -f |
5、以ROOT用户，在当前节点配置Oracle ASMFD。
|
[root@19crac1 ~]# asmcmd afd_configure ASMCMD-9521: AFD is already configured |
6、验证Oracle ASMFD的状态。
|
[root@19crac1 ~]# asmcmd afd_state ASMCMD-9526: The AFD state is 'LOADED' and filtering is 'ENABLED' on host '19crac1.oracle.com' [root@19crac1 ~]# |
7、以ROOT用户，借助步骤1中收集的数据，迁移所有UDEV磁盘。
|
[root@19crac1 disks]# asmcmd afd_label DG_DATA_0001 /dev/oracleasm/disks/mpathk --migrate [root@19crac1 disks]# asmcmd afd_label DG_DATA_0000 /dev/oracleasm/disks/mpathj --migrate [root@19crac1 disks]# asmcmd afd_label DG_DATA_0002 /dev/oracleasm/disks/mpathl --migrate [root@19crac1 disks]# asmcmd afd_label DG_GRID_0000 /dev/oracleasm/disks/mpathh --migrate [root@19crac1 disks]# asmcmd afd_label DG_GRID_0001 /dev/oracleasm/disks/mpathi --migrate |
8、验证所有磁盘都已迁移到ASMFD。
|
[root@19crac1 disks]# asmcmd afd_lsdsk -------------------------------------------------------------------------------- Label Filtering Path ================================================================================ DG_DATA_0000 ENABLED /dev/oracleasm/disks/mpathj DG_DATA_0001 ENABLED /dev/oracleasm/disks/mpathk DG_DATA_0002 ENABLED /dev/oracleasm/disks/mpathl DG_GRID_0000 ENABLED /dev/oracleasm/disks/mpathh DG_GRID_0001 ENABLED /dev/oracleasm/disks/mpathi [root@19crac1 disks]# |
9、重启ACFS和AFD，再次验证所有磁盘都已迁移到ASMFD。
|
[root@19crac1 disks]# acfsload stop [root@19crac1 disks]# afdload stop [root@19crac1 disks]# acfsload start [root@19crac1 disks]# afdload start [root@19crac1 disks]# asmcmd afd_lsdsk |
10、在其他的计算节点上执行上面的步骤（除了步骤7外）。
11、以GRID用户运行如下命令，更新ASM磁盘的搜索路径，将来ASM只识别ASMFD设备。
|
[grid@19crac1 ~]$ asmcmd dsset 'AFD:*' |
12、重启CRS集群，检验已经切换到AFD方式。
|
SQL> set lines 200 SQL> col path for a30 SQL> col name for a30 SQL> col header_status for a30 SQL> select name, path,header_status from v$asm_disk order by 1;
NAME PATH HEADER_STATUS ------------------------------ ------------------------------ ----------------------- DG_DATA_0000 AFD:DG_DATA_0000 MEMBER DG_DATA_0001 AFD:DG_DATA_0001 MEMBER DG_DATA_0002 AFD:DG_DATA_0002 MEMBER DG_GRID_0000 AFD:DG_GRID_0000 MEMBER DG_GRID_0001 AFD:DG_GRID_0001 MEMBER
SQL> |