Database Reference
In-Depth Information
Chapter 5
Storage and ASM Practices
by Kai Yu
In this chapter, I will discuss one of the key infrastructure components for the Real Application Cluster (RAC)
database: storage. In the RAC Database environment, storage needs to be accessible and shared by all the nodes
of the RAC Database. Database files such as data files, control files, online redo log files, temp files, spfiles, and
the Flash Recovery Area (FRA) are kept in this shared storage. In addition, two key sets of files of the Oracle
Clusterware—Oracle Cluster Registry (OCR) and voting disk files—are also stored here. Due to the presence of these
key components of the RAC Database and Clusterware, shared storage is one of the most critical components of the
RAC Database infrastructure.
This shared storage plays a key role in the stability of the Oracle Clusterware. In case any node of the RAC cluster
is not able to access the voting disk files that are stored in the storage within a predetermined time threshold (default
is 200 seconds), the RAC node will be evicted and get rebooted. As a result, the database instance on the RAC node
will also get rebooted. Even more seriously, if the OCR on the storage is lost permanently, the Oracle Clusterware
on the entire RAC cluster will no longer function. If this happens, the OCR needs to be recovered from its backup to
resume the normal operation of Oracle Clusterware.
Shared storage is also essential for the availability of the RAC Database. Any storage issue may lead to the loss of
a partial or entire RAC Database. For example, losing access to shared storage by a RAC node will bring down the RAC
Database instance on the RAC node. Loss of a data file or data corruption due to disk errors may cause the loss of a
partial or entire RAC Database.
The performance of the RAC Database depends upon the I/O performance of the shared storage. Storage I/O
performance is even more important in a RAC Database environment where the heavier I/Os from multiple RAC
Database nodes to a single shared storage may trigger huge I/O bottlenecks that hinder database performance.
As an essential part of the life cycle management of a RAC Database, the design, configuration, and management
of such a shared storage system is vital for the long-term health and performance of the RAC Database. However, in a
typical IT department, storage administration and database administration are two separate job responsibilities that
may belong to two different teams, in line with the separation of duties policy commonly followed in IT. Cooperation
and mutual understanding between the storage administrator and the database administrator (DBA) are crucial
for RAC Database storage design and configuration. In order to achieve database availability and performance
requirements defined by the database SLA (Service Level Agreement), some special design considerations and
best practices should be incorporated into the storage provisioning process for Oracle RAC Databases. Here we can
highlight some of the major tasks of storage provisioning and examine how the different roles such as the storage
administrator, the system administrator, and the DBA can play together.
