Wednesday, September 16, 2009

Oracle ACFS advanced topics

While everybody is exploring new features of the latest release of Oracle 11g, we're all very excited about all the new possibilities.
One of those new features is Oracle ACFS (ASM Cluster File System), wich provides an Oracle Home shared file system or a general purpose clustered file system on ASM storage. These are very nice features that will be used by many customers, in time to come. My first experiences are quite good. However, it is of great importance to know the limitations of those new features as well. Oracle has - on the topic of ACFS - thought about this very carefully, by adding an Appendix to the Storage Administrator's Guide, called ACFS Advanced Topics. Amongst these, a number of limitations are discussed. Well, one can argue whether these are actual limitations, that really limit you in operating ACFS, but it is good to know about these when planning for you Grid Infrastructure.

Oracle ACFS Disk Space Usage
Some Facts and Figures: Oracle ACFS supports:
  • 64 million files in a file system,
  • 63 snapshots,
  • 64 mounts on 32 bit systems,
  • 256 mounts on 64 bit systems.

Oracle allocates metadata for each node that mounts the file system, comprising approximately 64 – 128 MB per node. Also, some space is reserved for local bitmaps to reduce contention on the global bitmap in order to find free space. This can add up to 128MB per node. All of this space is reported as “in use” when querying the file system for disk space.

Oracle ACFS Error Handling
CAUTION: It is very important to unmount any file system that is using the Oracle ASM Dynamic Volume Manager (ADVM), before shutting down an Oracle ASM Instance. Failing to do so can result in I/O failures. One can never guarantee that the ASM instance(s) will never fail, but in case it does, volumes must be closed and opened again to be able to access them. Dismount the file systems that were mounted during failure and after the instance is restarted, mount the corresponding disk group with the volume enabled, after which you can mount the file system again.

Therefore, if any file systems are currently mounted on Oracle ADVM volume files, the SHUTDOWN ABORT command should not be used to terminate the Oracle ASM instance without first dismounting those file systems.

When a metadata write fails, regardless whether it be an ASM or storage failure, ACFS isolates the errors to a single file system and puts it in an off-line state. The only thing you can do with it is dismount. Other nodes (when available) recover the transaction, if it can write to that particular storage. After this, the file system can be mounted again.

When the file system being put offline is in use by processes or users, it might not be possible to unmount it. The administrator has to identify the processes that are claiming the file system and stop them before unmounting.

Oracle ACFS and NFS
Because block device major numbers of ADVM devices can be different after a reboot, the -fsid=# should be used when exporting NFS file systems. The option forces the clients to use this number (#) when communicating with this file system. Obviously, the number should be unique across the cluster. It replaces the number derived from the major/minor number of the device itself, since it can change with reboots.

Limits of Oracle ADVM
Some Facts: The default configuration for an Oracle ADVM volume is four columns of 64 MB extents in length and a 128 KB stripe width. Oracle ADVM writes data as 128 KB stripe chunks in round robin fashion to each column and fills a stripe set of four 64 MB extents with 2000 stripe chunks before moving to a second stripe set of four 64 MB extents for volumes greater than 256 megabytes. Note that setting the number of columns on an Oracle ADVM dynamic volume to 1 effectively turns off striping for the Oracle ADVM volume (are you still with me?;-)

On Linux platforms Oracle ASM Dynamic Volume Manager (Oracle ADVM) volume devices are created as block devices regardless of the configuration of the underlying storage in the Oracle ASM disk group. Do not use raw (8) to map Oracle ADVM volume block devices into raw volume devices.

Oracle ACFS and Oracle Restart
Oracle Restart does not support root-based Oracle ACFS resources for this release. As a result, the following operations are not automatically performed:
  • Loading Oracle ACFS drivers
  • Mounting Oracle ACFS file systems listed in the Oracle ACFS mount registry
  • Mounting resource-based Oracle ACFS database home file systems
The Oracle ACFS resources associated with these actions are not created for Oracle Restart configurations.
In short, the following resources are only supported for the Oracle Grid Infrastructure Cluster configurations, and are not supported for Oracle Restart Configurations:
  • The Oracle ACFS drivers resource (ora.drivers.acfs),
  • The Oracle ACFS registry resource (ora.registry.acfs)
  • The Oracle ACFS individual file system resource ora.diskgroup.volume.acfs)

Source: Oracle Database Storage Administrator's Guide, 11g Release 2, Appendix B

No comments:

Post a Comment