The following article will describe at a high-level the key ASM improvements in Oracle 12.2. Note, due to number of changes and improvements in ACFS, we cover ACFS is a separate article.
Oracle ASM Flex Disk Groups and File Groups
In previous releases disk group storage attributes were defined only at the disk group level, which was generally quite coarse from a storage management perspective. In 12.2, the concept of ASM flex disk group is introduced to enable users to manage storage at the database level, allowing greater granularity of control. Note, this capability is complimentary to the existing disk group manageability. Flex groups is built on the concept of File groups. File groups are a group of files that share the same set of properties and characteristics and used to describe database files. A significant benefit of file groups is the capability to have a different availability specification for each database; a key example is capability to create point-in-time database clones.
The following example illustrates how flex diskgroup is create.
SQL> create diskgroup vna_data flex redundancy disk ‘/dev/mapper/mpath*’
Or Convert existing standard diskgroup to flex diskgroup
SQL> alter diskgroup vna_data convert redundancy to flex;
Each database that is built on ASM (that has compatible.asm and compatible.rdbms is set to 220.127.116.11) will consist of file group, which allows granular storage management capabilities, such as redundancy, rebalance power limits and priority, striping and quota groups; i.e., at the file group level. Since the redundancy is now at the file group level and not at the disk group level, you can convert and change the redundancy of database from normal to high. Note, that you cannot currently change from external to high/normal.
Disk group can contain multiple file groups each with its independent Flex redundancy. A file group can belong to only one disk group; however, a database can span multiple disk groups with multiple file groups in different disk groups
ASM Support for Preferred Read on Extended Clusters
In previous releases the ASM_PREFERRED_READ_FAILURE_GROUPS defined read preference to a specific failure group in extended clusters.
In ASM 12.2, the preferred read failure groups capability is now automatically detected and set in the ASM instance when extended clusters are deployed. The ASM instance will evaluate which disks are local to that instance and set preference accordingly. Thus the ASM_PREFFERED_READ parameter is no longer necessary.
In 12.2 Oracle extends the capability of Flex ASM by enabling the ASM instance to be disjointed and separate from the physical servers hosting databases. This feature is referred to as ASM IO Server (IOS). This “far cluster” capability enables the deployment of larger clusters of ASM instances that can support more database clients while reducing the ASM instance footprint, thus storage consolidation occurs by placing larger number database into a single set of disk groups.
With the introduction of IOS, ASM and database storage access can be configured in the following configurations:
- direct access to ASM disks with local clients (same as pre-12.2)
- Flex ASM clients with direct access to ASM disks
- ACFS access through the ASM proxy instance
- Remote-Network based connectivity to ASM disk groups with Oracle IOServer (IOS)
Updates for Oracle ASM Filter Driver Installation and Configuration
Oracle 12.1 introduced ASM filter driver(ASMFD) for improved device management and disk group protection; however, the installation required upfront enablement before Grid Infrastructure. In 12.2, the installation and configuration for ASM Filter Driver (ASMFD) is now streamlined and enabled as part of the Oracle Grid Infrastructure installation.
ASM Extended Support for 4K Sector Size
The full support for 4K sector sizes has been looming since 11.2. Now in 12.2 ASM, a new diskgroup attribute, logical_sector_size, will define the logical sector size (in bytes) of the disk group and specifies the smallest I/O that can be issued to the underlying ASM disks.
Deprecation of Oracle ASM Intelligent Data Placement
Deprecation of ASM_PREFERRED_READ_FAILURE_GROUPS Initialization Parameter