Backup and Recovery
One common misconception about Oracle RAC is
that it provides all the backup necessary
because multiple nodes are in use.
However, it is important to remember that
this redundancy is only at the instance level,
not the database level. This means that if a
failure occurs at the database level, such as
lost or corrupt datafiles, RAC will not be able
to keep the system online by itself.
There are two types of failure that must be
considered:
-
Database Failure
is caused by a loss or corruption at the
file level.
Remember that while instances are
redundant, database files are not.
-
Instance Failure
is caused when an instance crashes.
In a RAC environment, the other
instances in the cluster can continue to
work, and the instance will recover itself
when the system comes back online.
RAC Backup and Recovery
While RAC is a complex system involving multiple
machines plugging into a centralized database,
backup of the environment is relatively simple.
The same basic tools or commands that are
used to backup a single instance database can,
in most cases, back up a RAC database.
A RAC database has one single set of datafiles
and control files with redo log groups for each
instance.
All of these files are centralized, as
RAC requires shared storage in order to run.
Backing up those centralized files is the
same as backing up files for a single instance,
except that there will be a few more files.
A DBA can use many methods to backup his RAC
database.
These methods include:
-
Logical export via
exp
or the Datapump
tool
expdp
-
User-created cold or hot backups
-
RMAN backups (hot or cold)
Before these backup options can be explored, it
is important to understand how RAC and backups
come together to form a complete redundancy
solution.
Maximum Availability Architecture (MAA)
Oracle recommends a standard known as the
Maximum Availability Architecture,
or MAA.
This plan provides contingencies both at
the instance level and the database level.
High Availability (HA)
Generally speaking, instance level redundancy is
known as High Availability (HA).
This means that if an instance is lost in
a RAC cluster, the high availability features
ensure that other instances will take over the
workload as seamlessly as possible.
HA does not cover datafile loss or
corruption.
Instead, it only covers the runtime
components of the Oracle environment.
High availability can be achieved in a RAC
environment when proper failover techniques are
used, such as Transparent Application Failover
(TAF) and/or Fast Connect Failover (FCF).
Though an instance failure will require
reconfiguration, i.e. instance recovery of the
lost instance, the basic components of RAC
should still be available.
