Traditionally, Oracle RAC is used in a multi-node
architecture, with many separate instances of RAC residing on separate
servers. This protects against unplanned server outages because
transparent application failover will re-direct to another server, and also
aids the RAC goal of scalability, when a new server instance is genned-in to
the RAC cluster.
Now in Oracle 11g r2, we see a new feature dubbed "RAC
One Node". RAC One Node claims to be a multiple instances of RAC
running on a single node in a cluster, and has a fast "instance relocation"
feature in cases of catastrophic server failure.
Both RAC and RAC One Node protect against server and
hardware failure (except disk failure), but they do it in different ways:
- RAC failover: RAC uses multiple
servers (nodes) and the tnsnames.ora file directs failover to a surviving
server when a node crashes.
- RAC One Node failover: RAC
One Node uses "instance relocation", and when an instance fails, RAC One
Node re-starts a failed instance on another node, by re-mounting the disk on
the new server and using the pfile/spfile to re-start the instance.
This instance relocation uses a new featured dubbed
Oracle Omotion.
While RAC One node will not protect you in a case of
server failure (unlike regular RAC, running nodes on many servers), RAC One
Node does offer instance duplication within a single server environment.
Oracle Corporation is now embracing the IT industry
concept of "instance consolidation", a movement to collected
together instances from the bad old days of client server computing where we
had a one instance/one server architecture. This is a Back to the
Future approach, a movement back to the monolithic server environments of
the 1980's, with all of their benefits:
-
Centralized patching and software maintenance.
-
Redundant instances mean zero outages during
patches and upgrades.
-
On-demand RAM and CPU sharing within the single
large server.
-
Less Oracle DBA resources required to manage dozens
of instances.
See my notes here on
Oracle
server consolidation and the 2nd age of mainframe computing.
RAC One Node and VMware
Oracle says that RAC One Node is useful in VMware
environments. For details on using Oracle on VMware see, the book "Oracle
on VMware: expert tips for Oracle virtualization" by Bert Scalzo.
"Using Oracle RAC or Oracle RAC One Node with Oracle VM
increases the benefit of Oracle VM
with the high availability and scalability of Oracle RAC.
If your VM is sized too small, you can migrate the Oracle
RAC One instance to another larger Oracle VM node in the cluster (using
Omotion) or move the Oracle RAC One instance to another Oracle VM node, and
then resize the Oracle VM.
When you move the Oracle RAC One Node instance to the
newly resized Oracle VM node, you can dynamically increase any limits
programmed with Resource Manger Instance Caging."
Fast Recovery vs. Continuous availability
In RAC we achieve continuous availability because that
is little or no latency as TAF redirects transactions to surviving server
when a server crashes or a node become unavailable. What makes Oracle
RAC One Node special is that it performs quick clone of the instance to a
new node:
"If the node fails, Oracle Clusterware will automatically
restart the Oracle RAC One Node instance on another server in the cluster."
This is a similar approach to instance relocation that
was first introduced by Savantis Systems with their
DB-Switch
invention, an offshoot of the Database Area Network (DAN)
approach.
Instance relocation for Oracle using Omotion
To understand how RAC One Node might do a fast instance relocation, lets
examine similar offerings by 3rd party vendors. Savantis Systems uses
a high-speed database switch (dbSwitch™) which is placed between the
application layer and the database layer.
This database switch enables Oracle
databases to be transparently relocated onto new servers without the need to
modify applications connections to the database. The database servers, in
turn, are connected to a shared storage such as SAN, such that database
files can be quickly remapped onto a different server.
In Oracle RAC One Node, it appears that the Omotion
software component uses VMware for the high speed instance relocation.
See here for details on
how Oracle instance relocation works using DAN and SAN technology.
Virtualization allows for instance relocation between servers
Here is how instance relocation worked in earlier offerings by 3rd parties.
Using IP address masks enables virtualization of database addresses. To the
database clients, each database is identified by a unique virtual IP address
(VIP). These VIPs are not defined on any specific machine on the network,
and are routable by the dbSwitch™ using Network Address Translation (NAT).
This allows the Database Area Network to move Oracle instances without
affecting the application server layer.
The Database Area Network (DAN) architecture.
The internal mechanism in Omotion for fast Oracle instance relocation may be
quite complex, but the idea is simple. In the olden days, we did this with a
SAN environment, using the following steps:
1. The Oracle instance or the server crashes.
2. A
dead connection probe (DCD) detects the outage and we un-mount the SAN data
file systems from the source server (if it's awake).
3. We then mount the database files onto the failover
server using SAN.
4. We re-start the instance on the new server,
using a pre-loaded init.ora file.
5 We use a failover routing
mechanism like transparent application failover (TAF) to redirect
transactions accessing the database (VIP) to the new server.
A typical time for an Oracle instance relocation is less than 20 seconds,
not true continuous availability, but certainly better than a catastrophic
unplanned outage.
Again, Oracle has not published the internal machinations
of their instance relocation strategy, but it's a safe bet that it happens
very quickly over SAN storage.
Oracle RAC One Node Utilities
Oracle RAC One Node introduces several new utilities:
- Oracle Database Resource Manager Instance Caging -
The new instance caging features is part of the database resource
manager in RAC One Node, a tool to allocate server resources to multiple
instances running on the same server.
- Omotion online migration utility - This is a new
RAC One Node utility that allows for a RAC instance to be quickly
"lifted" and copied to a new server node.
In sum, Oracle RAC One Node is a solution aimed at
smaller shops and systems that need fast failover, not not necessarily 100%
continuous availability. Oracle RAC One Node also buys in to the
industry-wide server consolidation movement, whereby instances are
consolidated for better quality management and resource sharing.
The hardware industry is also embracing this "scale up"
approach by providing servers with lots of slots for future expansion.
For example, the Sun Fire X4640 Server, allows from 12 to 48 processors and
up to 512 gig RAM, allowing the DBA to add server resources just-in-time to
meet increases in processing demands.
