Using Oracle ACFS Replication

Oracle ASM Cluster File System (ACFS) replication is a new feature in 11.2.0.2 on Linux; as its name suggests, it allows you to replicate files from one host to another. The terminology used is reminiscent of Oracle Data Guard, with a primary file system and a standby file system, with replication logs recording the changes. The replication logs are shipped across the network from the primary to the standby and the changes are applied.

When considering ACFS replication, your most important concern is probably file system space usage. Replication logs take up space in the file system being replicated; if you cannot complete a replication, you cannot complete a change on the file system, so running out of space in this setup can cause major problems. The standby must also not run out of space because that would stop the application of changes on the standby and cause replication logs to pile up on the primary, because they are deleted from both sites only when they have been applied.

You must also consider the network bandwidth to the standby, because slow transfer of logs will also mean that the standby will be out of date and logs will remain on the primary for longer than needed. Finally, the standby needs to be up to the job of applying the logs as fast as the primary generates them, so having a significantly lower powered standby machine is not a good idea.

Because the ability to determine the rate of change appears to be so important when sizing the file system, network connection, and standby, Oracle has provided a way to determine how to use the acfsutil info fs command. (Note that the 5 in the first line is the period in seconds we are reporting for each time we write the amount of change. The command will continue reporting statistics every 5 seconds until you break out of it.)

   [grid]$ /sbin/acfsutil info fs -s 5 /u02/app/oracle/acfsdbhome1/
   /u02/app/oracle/acfsdbhome1/
       amount of change since mount:       9851.31 MB
       average rate of change since mount: 154 KB/s

       amount of change: 0.50 MB   rate of change: 102 KB/s
       amount of change: 3.00 MB   rate of change: 612 KB/s
       amount of change: 1.00 MB   rate of change: 204 KB/s

We will not cover the calculations here to make sure you have enough resources to support replication. These can be found in the Oracle Automatic Storage Management Administrator’s Guide.

Following are other considerations for ACFS replication:

• A server can be both a replication primary and standby system for different file systems.
• Each primary file system can have only one standby file system.
• You can replicate a file system only if it is mounted by eight or fewer nodes.
• Replication works on single-instance and clustered systems.
• ACFS tagging can be used to replicate only certain files within a file system, rather than the whole file system.
• File systems that are encrypted or under ACFS security control cannot be replicated.
• The standby file system must be running on a server that runs the same operating system and Oracle software distribution as the primary.
• The files on the standby are always available for reading, but they may change at any time.