Oracle RAC Database Deployment Models (Changes in 26ai & earlier)
Starting
with Oracle Database 21c, there is a single, merged management style for Oracle
RAC databases.
Note:
The
policy-managed database deployment option is de-supported in Oracle Database 26ai.
Prior to
Oracle Database 21c, Oracle RAC databases support two different management
styles and deployment models: administrator-managed deployment and
policy-managed deployment.
·
Administrator-managed
deployment requires
that you statically configure each database instance to run on a specific node
in the cluster. This deployment also requires that you configure database
services to run on specific instances that belong to a particular database using
the preferred and available designation.
·
Policy-managed
deployment is
based on server pools. In this deployment, database services run within a
server pool as singleton or uniform across all of the servers in the server
pool.
Databases
are deployed in one or more server pools and the size of the server pools
determine
the number of database instances in the deployment.
Starting
with Oracle Database 21c, the two management styles are merged into a single
deployment model that combines the best features of each model. The administrator-managed
database deployment style now has additional capabilities that were previously
available only in policy-managed databases.
These
enhancements result in a new, converged deployment style. To use the merged
database management style, you must have a Container Database
(CDB) with
at least one Pluggable Database (PDB).
You manage
the merged database deployment model using the same commands and utilities (such
as SRVCTL, Oracle DBCA, or Oracle Enterprise Manager) as before. All commands
and utilities, except for the policy-management specific commands such as
srvpool commands, maintain backward compatibility to support the management of
Oracle databases prior to Oracle Database 21c.
The merged
database management style simplifies management of dynamic systems. The clusters
and databases can expand or shrink as requirements change. The Oracle home software
must be installed on every node in the cluster.
Benefits of the merged management style
The PDBs in
the cluster database are available on all nodes, or a subset of the nodes,
based on the cardinality setting for the PDB. The cardinality of a PDB
governs the number of nodes where a PDB can run at the same time. If you use a
number for cardinality instead of ALL, then the instances in which the PDBs are
opened depend on the resources available to each instance.
A database
instance is started on every server in the cluster that hosts a PDB. If you are
using Oracle Automatic Storage Management (Oracle ASM) with Oracle Managed
Files for your database storage, then, when an instance starts and there is no
redo thread available, Oracle RAC automatically enables one and creates the
required redo log files and undo tablespace.
Clients can
connect to a PDB using the same SCAN-based connect string irrespective of which
servers the PDBs are running on at the time.
What Was “Policy-Managed Database Deployment”?
In Oracle
RAC environments (11gR2 through 19c), you could configure your database in two
deployment modes:
|
Mode |
Description |
Node Association |
Tools / Features |
|
Administrator-Managed |
You manually specify which
instances run on which nodes. |
Fixed to specific nodes (e.g.,
node1, node2). |
Classic mode, simple for small
clusters. |
|
Policy-Managed |
You define server pools
(logical groups of servers), and Oracle Clusterware automatically places and
relocates instances. |
Dynamic — instances float based on
pool policies. |
Used by QoS Management, Instance
Caging, and server pools. |
Instead, you defined:
- Server Pools: named groups with MIN_SIZE, MAX_SIZE, and
IMPORTANCE.
- Databases and Services assigned to those pools.
- QoS Management Policies operating on those pools.
Clusterware dynamically placed or moved instances between servers in those pools depending on workload, resource availability, or QoS recommendations.
Example:
srvctl
add serverpool -serverpool oltp_pool -min 2 -max 4 -importance 100
srvctl
modify database -db FINDB -serverpool oltp_pool
This was the
policy-managed model ; the foundation of Oracle QoS Management in
RAC.
Oracle
officially desupported policy-managed database deployments in Oracle
Database 26ai (23c).
That means:
- You can no longer create or use policy-managed
databases.
- Server pools and their automatic instance
placement model are removed.
- RAC databases must now use administrator-managed
configuration.
What “Desupported” Means
“Desupported”
in Oracle terminology means:
- You cannot create new policy-managed databases in
26ai.
- Existing policy-managed databases must be
converted to administrator-managed before upgrade.
- The related commands (like srvctl add serverpool,
srvctl modify database -serverpool, etc.) are no longer valid for database
placement.
- QoS Management functionality that depended on
server pools is no longer available.
Why Oracle Made This Change
There are
several reasons Oracle deprecated this model:
1.
Complexity
Many customers found server pool and policy-managed setups confusing to operate
and troubleshoot.
2.
Low Adoption
Most real-world RAC
environments (especially Exadata and production workloads) used
administrator-managed mode.
3.
Integration Simplification
Oracle wanted to streamline cluster management around a single, predictable
model — administrator-managed.
4.
Shift Toward Autonomous and Multitenant
Oracle’s new architectures (Autonomous Database, Multitenant, Cloud RAC) handle
resource control differently, without manual server pool constructs.
What Happens to QoS Management
QoS
Management in its classic form (policy-based RAC control via server pools)
is effectively retired starting with Oracle 26ai.
Key
implications:
- QoS Management relied on server pools and
performance classes; these no longer exist in 26ai.
- QoS metrics and recommendations (via Enterprise Manager or
EMCLI) are not available in 26ai RAC.
- Resource isolation and
prioritization should now be handled using:
- Oracle Resource Manager
- Instance Caging
- CDB/PDB Resource Plans
- Services and Load Balancing
Advisory
So, in 26ai
and future versions, QoS-style behavior is expected to move to Multitenant
resource management and Autonomous Database infrastructure, not
Clusterware-level QoS.
Before
upgrading to 26ai:
1.
Convert your policy-managed databases to
administrator-managed.
srvctl convert database -db FINDB -admin-managed
2.
Retire server pool usage; manage instances explicitly on
specific nodes.
3.
Migrate QoS controls to:
o Oracle Resource
Manager plans (CPU, I/O, session control)
o Service-level
resource directives
o Instance Caging
for CPU isolation
Oracle
provides migration utilities and step-by-step guides in:
Oracle
Database 26ai Upgrade Guide → Desupported Features → Policy-Managed Database
Deployments
Summary
|
Feature |
Status in 26ai |
Replacement |
|
Policy-managed databases |
Desupported |
Administrator-managed |
|
Server pools |
Desupported |
Explicit node instance mapping |
|
Oracle QoS Management |
Retired |
Resource Manager / Instance Caging |
|
Dynamic instance placement |
Removed |
Manual instance placement |
|
RAC QoS metrics |
Unavailable |
EM performance monitoring /
Resource plans |
Oracle
Database 26ai eliminates the policy-managed
deployment model and its dependent technologies (like QoS Management based on
server pools).
RAC databases must now be deployed in administrator-managed mode,
with performance and isolation handled via Resource Manager and instance
caging, not server pool reallocation.
++++++ Alireza Kamrani
++++++
No comments:
Post a Comment