online documentation (CCVU.HLP) included on the CD.
Compaq Insight Manager
Compaq Insight Manager, loaded from the Compaq Management CD that is
shipped with each ProLiant server, is an easy-to-use, console-based software
utility for collecting server and cluster information. Compaq Insight Manager
performs the following functions:
I Monitors fault conditions and system status
I Monitors shared storage and interconnect adapters
I Forwards server alert fault conditions
I Remotely controls servers
The Integrated Management Log collects and feeds data to Compaq Insight
Manager. This log is used with the Insight Management Desktop (IMD),
Remote Insight (optional controller), and SmartStart.
In Compaq servers, each hardware subsystem, such as disk storage, system
memory, and system processor, has a robust set of management capabilities.
Compaq Full Spectrum Fault Management notifies of impending fault
conditions and keeps the server up and running in the unlikely event of a
hardware failure.
For information concerning Compaq Insight Manager, refer to the Compaq
Server Setup and Management pack shipped with each ProLiant server.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: b-ch1 Architecture of the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 11:58
AM
1-20 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Compaq Insight Manager XE
Compaq Insight Manager XE is a Web-based management system and is
located on the Compaq Management CD shipped with each ProLiant server. It
can be used in conjunction with Compaq Insight Manager agents as well as its
own Web-enabled agents. This browser-based utility provides increased
flexibility and efficiency for the administrator. It extends the functionality of
Compaq Insight Manager and works in conjunction with the Cluster Monitor
subsystem, providing a common data repository and control point for
enterprise servers and clusters, desktops, and other devices using either
SNMP- or DMI-based messaging.
Cluster Monitor
Cluster Monitor is a Web-based monitoring subsystem of Compaq Insight
Manager XE. With Cluster Monitor, you can view all clusters from a single
browser and configure monitor points and specific operational performance
thresholds that will alert you when these thresholds have been met or exceeded
on your application systems. Cluster Monitor relies heavily on the Compaq
Insight Manager agents for basic information about system health. It also has
custom agents that are designed specifically for monitoring cluster health.
Cluster Monitor provides access to the Compaq Insight Manager alarm,
device, and configuration information.
Cluster Monitor allows the administrator to view some or all of the clusters,
depending on administrative controls that are specified when clusters are
discovered by Compaq Insight Manager XE.
Compaq Intelligent Cluster Administrator
Compaq Intelligent Cluster Administrator extends Compaq Insight Manager
and Cluster Monitor by enabling Administrator to configure and manage
ProLiant clusters from a Web browser. With Compaq Intelligent Cluster
Administrator, you can copy, modify, and dynamically install a cluster
configuration on the same physical cluster or on any physical cluster anywhere
in the system, through the Web.
Compaq Intelligent Cluster Administrator checks for any cluster destabilizing
conditions, such as disk thresholds or application slowdowns, and reallocates
cluster resources to meet processing demands. This software also performs
dynamic allocation of cluster resources that may be failing without causing the
cluster to fail over.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: b-ch1 Architecture of the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 11:58
AM
Architecture of the Compaq ProLiant Clusters HA/F100 and HA/F200 1-21
Compaq Intelligent Cluster Administrator also provides initialized cluster
configurations that allow rapid cluster generation as well as cluster
configuration builder wizards for extending the Compaq initialized
configurations.
Compaq Intelligent Cluster Administrator is included with the
HA/F200 cluster kit and can be purchased as a stand-alone component for the
HA/F100 cluster. Intelligent Cluster Administrator is licensed on a per cluster
basis.
Resources for Application Installation
The client/server software applications are among the key components of any
cluster. Compaq is working with its key software partners to ensure that
cluster-aware applications are available and that the applications work
seamlessly on Compaq ProLiant clusters.
Compaq provides a number of Integration TechNotes and White Papers to
assist you with installing these applications in a Compaq ProLiant Cluster
environment.
Visit the Compaq High Availability website
(http://www.compaq.com/highavailability) to download current versions of these
TechNotes and other technical documents.
IMPORTANT: Your software applications may need to be updated to take full advantage
of clustering. Contact your software vendors to check whether their software supports
MSCS and to ask whether any patches or updates are available for MSCS operation.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: b-ch1 Architecture of the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 11:58
AM
2
Chapter
Designing the Compaq ProLiant
Clusters HA/F100 and HA/F200
Before connecting any cables or powering up any machines, it is important to
understand how all of the cluster components and concepts fit together to meet
your information system needs. The major topics discussed in this chapter are:
I Planning Considerations
I Capacity Planning
I Network Considerations
I Failover/Failback Planning
In addition to reading this chapter, read the planning chapter in Microsoft
documentation that came with your operating system.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-2 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Planning Considerations
To correctly assess capacity, network, and failover needs in your business
environment, it is important to understand clustering and the things that affect
the availability of clusters. The items detailed in this section will help you
design your Compaq ProLiant Cluster so that it addresses your specific
availability needs.
I Cluster configuration design is addressed in "Cluster Configurations."
I A step-by-step approach to creating cluster groups is discussed in
"Cluster Groups."
I Recommendations regarding how to reduce or eliminate single points of
failure are contained in the "Reducing Single Points of Failure in the
HA/F100 Configuration" section of this chapter. By definition, a highly
available system is not continuously available and therefore may have
single points of failure.
NOTE: The discussion in this chapter relating to single points of failure applies only to the
Compaq ProLiant Cluster HA/F100. The HA/F200 includes dual redundant loops, that
eliminate certain single points of failure contained in the HA/F100.
Cluster Configurations
Although there are many ways to set up clusters, most configurations fall into
two categories: active/active and active/standby.
Active/Active Configuration
The core definition of an active/active configuration is that each node is
actively processing data when the cluster is in a normal operating state. Both
the first and second nodes are "active." Because both nodes are processing
client requests, an active/active design maximizes the use of all hardware in
both nodes.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-3
An active/active configuration has two primary designs:
I The first design uses MSCS failover capabilities on both nodes, enabling
Node 1 to fail over clustered applications to Node 2 and enabling Node
2 to fail over clustered applications to Node 1. This design optimizes
availability since both nodes can fail over applications to each other.
I The second design is a one-way failover. For example, the Microsoft
clustering software may be set up to allow Node 1 to fail over clustered
applications to Node 2, but not to allow Node 2 to fail over clustered
applications to Node 1. While this design increases availability, it does
not maximize availability since failover is configured on only one node.
When designing cluster nodes to fail over to each other, ensure that each
server has enough capacity, memory, and processor power to run all
applications (all applications running on the first node plus all clustered
applications running on the other node).
When designing your cluster so that only one node (Node 1) fails over to the
other (Node 2), ensure that Node 2 has enough capacity, memory, and CPU
power to execute not only its own applications, but to run the clustered
applications that can fail over from Node 1.
Another consideration when determining your servers' hardware is
understanding your clustered applications' required level of performance when
the cluster is in a degraded state (when one or more clustered applications is
running on a secondary node). If Node 2 is running near peak performance
when the cluster is in a normal operating state, and if several clustered
applications are failed over from Node 1, Node 2 will likely execute the
clustered applications more slowly than when they were executed on Node 1.
Some level of performance degradation may be acceptable. Determining how
much degradation is acceptable depends on the company.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-4 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Example 1: File & Print/File & Print
An example business scenario (Figure 2-1) involves two file and print servers.
The Human Resources (HR) department uses one server, and the Marketing
department uses the other. Both servers actively run their own file shares and
print spoolers while the cluster is in its normal state (an active/active design).
If the HR server encounters a failure, it fails over its file and print services to
the Marketing server. HR clients experience a slight disruption of service
while the file shares and print spooler fail over to their secondary server. Any
jobs that were in the print spooler before the failure event will now print from
the Marketing server.
File and Print File and Print
Marketing Human Resources
Capacity Capacity
Human Resources Marketing
Shared Storage
(Marketing) (Human Resources)
Figure 2-1. Active/active example 1
When failover is complete, all of the HR clients have full access to their file
shares and print spooler. Marketing clients do not experience any disruption of
service. All clients may experience slowed performance while the cluster runs
in a degraded state.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-5
Example 2: Database/Database
Another scenario (Figure 2-2) has two distinct database applications running
on two separate cluster nodes. One database application maintains Human
Resources records, and its primary node is set to the HR database node. The
other database application is used for market research, and its primary node is
set to the Marketing database node.
Order Entry Order Entry
Database Database
Shared Storage
Node 1 Node 2
(Order Entry) (Order Entry)
Figure 2-2. Active/active example 2
While in a normal state, both cluster nodes run at expected performance levels.
If the Marketing server encounters a failure, the market research application
and associated data resources fail over to their secondary node, the HR
database server. The Marketing clients experience a slight disruption of
service while the database resources are failed over, the database transaction
log is rolled back, and the information in the database is validated. When the
database validation is complete, the market research application is brought
online on the HR database node and the Marketing clients can reconnect to it.
While the Marketing database validation is occurring, the HR clients do not
experience any disruption of service.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-6 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Example 3: File & Print/Database
In this example (Figure 2-3), a business uses a single server to run its order
entry department. The same department has a file and print server. While order
entry is business-critical and requires maximum availability, the file and print
server can be unavailable for several hours without impacting revenue. In this
scenario, the order entry database is configured to use the file and print server
as its secondary node. However, the file and print server will not be configured
to fail over applications to the order entry server.
File and Print
Order Entry
Services
Database
Capacity of
Order Entry
Shared Storage Database
Node1 Node2
(File and Print)
(Order Entry)
Figure 2-3. Active/active example 3
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-7
If the node running the order entry database encounters a failure, the database
fails over to its secondary node. The order entry clients experience a slight
disruption of service while the database resources are failed over, the database
transaction log is rolled back, and the information in the database is validated.
When the database validation is complete, the order entry application is
brought online on the file and print server and the clients can reconnect to it.
While the database validation is occurring, file and print activities continue
without disruption.
If the file and print server encounters a failure, those services are not failed
over to the order entry server. File and print services are offline until the
problem is resolved and the node is brought back online.
Active/Standby Configuration
The primary difference between an active/active configuration and an
active/standby configuration is the number of servers actively processing data.
In active/standby, only one server is processing data (active) while the other
(the standby server) is in an idle state.
The standby server must be logged in to the Windows NT or Windows 2000
domain and the Microsoft clustering software must be up and running.
However, no applications are running. The standby server's only purpose is to
take over failed clustered applications from its partner. The standby server is
not a preferred node for any clustered applications and, therefore, does not fail
over any applications to its partner server.
Because the standby server does not process data until it accepts failed over
applications, the limited use of the server may not justify the cost of the server.
However, the cost of standby servers is justified when performance and
availability are paramount to a business' operations.
The standby server should be designed to run all of the clustered applications
with little or no performance degradation. Since the standby server is not
running any applications while the cluster is in a normal operating state, a
failed-over clustered application will likely execute with the same speed and
response time as if it were executing on the primary server.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-8 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Example 4: Database/Standby Server
An example business scenario describes a mail order business whose
competitive edge is quick product delivery (Figure 2-4). If the product is not
delivered on time, the order is void and the sale is terminated. The business
uses a single server to perform queries and calculations on order entry
information, translating sales orders into packaging and distribution
instructions for the warehouse. With an estimated downtime cost of
$1,000/hour, the company determines that the cost of a standby server is
justified.
This mission-critical (active) server is clustered with a standby server. If the
active server encounters a failure, this critical application and all its resources
fail over to the standby server, which validates the database and brings it
online. The standby server now becomes active and the application executes at
an acceptable level of performance.
Capacity
Mail Order System
(Mail Order System)
Shared Storage
Node1 Node2
(Standby)
(Mail Order Database)
Figure 2-4. Active/standby server example
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-9
Cluster Groups
Understanding the relationship between your company's business functions
and cluster groups is essential to getting the most from your cluster. Business
functions rely on computer systems to support activities such as transaction
processing, information distribution, and information retrieval. Each computer
activity relies on applications or services, and each application depends on
software and hardware subsystems. For example, most applications need a
storage subsystem to hold their data files.
This section is designed to help you understand which subsystems, or
resources, must be available for either cluster node to run a clustered
application properly.
Creating a Cluster Group
The easiest approach to creating a cluster group is to start by designing a
resource dependency tree. A resource dependency tree has as its top level the
business function for which cluster groups are created. Each cluster group has
branches that indicate the resources upon which the group is dependent.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-10 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Resource Dependency Tree
The following steps describe the process of creating a resource dependency
tree. Each step is illustrated by adding information to a sample resource
dependency tree. The sample is for a hypothetical Web Sales Order business
function, which consists of two cluster groups: a database server
(a Windows NT or Windows 2000 application) and a Web server (a
Windows NT or Windows 2000 service).
NOTE: For this example, it is assumed that each cluster group can communicate with the
other even if they are not executing on the same node, for example, by means of an IP
address. With this assumption, one cluster group can fail over to the other node, while the
remaining cluster group continues to execute on its primary node.
1. List each business function that requires a clustered application
or service (Figure 2-5).
Web Sales Order
Business Function
Web Sales Order
Cluster Group
Cluster Group #2
Cluster Group #1
Figure 2-5. Resource dependency tree: step 1
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-11
2. List each application or service required for each business function
(Figure 2-6).
Web Sales Order
Business Function
Web Server Service Database Server Application
(Cluster Group #1) (Cluster Group #2)
Resource Resource Resource Resource Resource Resource Resource
#1 #2 #3 #1 #2 #3 #4
Dependent-Resource Dependent-Resource
#1 #1
Figure 2-6. Resource dependency tree: step 2
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-12 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
3. List the immediate dependencies for each application or service (Figure
2-7.
Web Sales Order
Business Function
Database Server Application
Web Server Service
(Cluster Group #2)
(Cluster Group #1)
Physical Disk
Web Server Physical Disk Physical Disk Database
Network Network
Resource -
Service Resource- Resource - Application
Name Name
contains DB
contains web contains DB
data file(s)
pages and web log file(s)
scripts
IP Address IP Address
Figure 2-7. Resource dependency tree: step 3
4. Transfer the resource dependency tree into a Cluster Group Definition
worksheet.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-13
Figure 2-8 illustrates the worksheet for the Web Sales Order business function.
A blank copy of the worksheet is provided in Appendix A.
Cluster Group Definition Worksheet
Web Sales Order
Cluster Function
Web Server Service
Group #1
Database Server Application
Group #2
Resource Definitions
Group #1 (Web Server Service)
Resource #1 Network Name
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
IP Address
Resource #2 Physical Disk Resource-contains Web pages and Web scripts
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
Resource #3 Web Server Service
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
Resource #4 N/A
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
Group #2 (Database Server Application)
Resource #1 Network Name
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
IP Address
Resource #2 Physical Disk Resource-contains database log files
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
Resource #3 Physical Disk Resource-contains database data files
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
Resource #4 Database Application
Sub Resource 1 Sub Resource 2 Sub Resource 3 Sub Resource 4
Figure 2-8. Cluster Group Definition Worksheet (example)
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-14 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Use the resource dependency tree concept to review your company's
availability needs. It is a useful exercise, directing you to record the exact
design and definition of each cluster group.
Reducing Single Points of Failure in the
HA/F100 Configuration
The final planning consideration is reducing single points of failure.
Depending on your needs, you may leave all vulnerable areas alone, accepting
the risk associated with a potential failure. Or, if the risk of failure is
unacceptable for a given area, you may elect to use a redundant component to
minimize, or remove, the single point of failure.
NOTE: Although not specifically covered in this section, redundant server components
(such as power supplies and processor modules) should be used wherever possible.
These features will vary based upon your specific server model.
The single points of failure described in this section are:
I Cluster interconnect
I Fibre Channel data paths
I Non-shared disk drives
I Shared disk drives
NOTE: The Compaq ProLiant Cluster HA/F200 addresses the single points of failure listed
above with its dual redundant loop configuration. For more information, refer to the
"Enhanced High Availability Features of the HA/F200" section of this chapter.
Cluster Interconnect
The interconnect is the primary means for the cluster nodes to communicate.
Intracluster communication is crucial to the health of the cluster. If
communication between the cluster nodes ceases, the Microsoft clustering
software must determine the state of the cluster and take action, in most cases
bringing the cluster groups offline on one of the nodes and failing over all
cluster groups to the other node.
Following are two strategies for increasing the availability of intracluster
communication. Combined, these strategies provide even more redundancy.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-15
Microsoft clustering software configuration
Microsoft Cluster Server for Windows NTS/E and Cluster Service for
Windows 2000 Advanced Server (MSCS) allow you to configure a primary
and backup path for intracluster communication, which will reduce the
possibility of an intracluster communication disruption. Any network interface
card (NIC) in the nodes can be configured to serve as a backup path for node-
to-node communication. When the primary path is disrupted, the transfer of
communication responsibilities goes undetected by applications running on the
cluster. Whether a dedicated or public interconnect has been set up, a separate
NIC should be configured to act as a redundant interconnect. This is an easy
and inexpensive way to add redundancy to intracluster communication.
Redundant Interconnect Card
Another strategy to increase availability is to use a redundant interconnect
card. This may be done for either the dedicated intracluster communication
path, or for the client LAN. If you are using a dedicated, direct-connection
interconnect configuration, you can install a second dedicated,
direct-connection interconnect.
NOTE: If you are using the ServerNet option as the interconnect, the card itself has a
built-in level of redundancy. Each ServerNet PCI adapter has two data ports, thereby
allowing two separate cables to be run to and from each cluster node. If the ServerNet
adapter determines that data is being sent from one adapter but not received by the
other, it will automatically route the information through its other port.
There are two implementations that provide identical redundant NIC
capability. The implementation you choose will depend on your hardware. The
Compaq TLAN Teaming and Configuration Utility is supported on all
Compaq TI-based Ethernet and Fast Ethernet NICs, such as NetFlex-3 and
Netelligent 10/100 TX PCI Ethernet NICs. The Compaq Network Teaming
and Configuration Utility is designed to operate with the Compaq Intel-based
10/100 NICs. Combining these utilities with the appropriate NICs will enable
a seamless, undetectable failover of the primary interconnect to the redundant
interconnect.
NOTE: These two methods of NIC redundancy cannot be combined in a single redundant
NIC pair: TI-based NICs may not be paired with Intel-based NICs to create a redundant
pair. For more information, refer to the Compaq White Paper, "High Availability Options
Supported by Compaq Network Interface Controllers," available at the Compaq High
Availability website (http://www.compaq.com/).
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-16 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Because the purpose of the redundant interconnect is to increase the
availability of the cluster, it is important to monitor the status of your
redundant NICs. Compaq Insight Manager and Compaq Insight Manager XE
simplify management of the interconnect by monitoring the state of the NIC.
You can view status information and alert conditions for all cards in each
node. If a failover event occurs due to a disruption in the heartbeat, you can
use the Compaq Insight Manager tools to determine where the disruption
originated.
Cluster-to-LAN Communication
Each cluster node must have at least one NIC that connects to the LAN.
Through this connection, network clients can access applications and data on
the cluster. If the LAN NIC fails in one of the nodes, any clients connected
directly to the cluster node by means of the computer name, cluster node
IP address, or MAC address of the NIC will no longer have access to their
applications. Clients connected to a virtual server on the cluster (via the
IP address or network name of a cluster group) reconnect to the cluster
through the surviving cluster node.
Failure of a LAN NIC in a cluster node may have serious repercussions.
If your cluster is configured with a dedicated interconnect and a single
LAN NIC, the failure of a LAN NIC will prevent network clients from
accessing cluster groups running on that node. If the interconnect path is not
disrupted, it is possible that a failover will not occur. The applications will
continue to run on the node with the failed NIC; however, clients will be
unable to access them.
Install redundant NICs and use the proper redundant NIC utility to reduce the
possibility of LAN NIC failure. When your cluster nodes are configured with
the utility, the redundant NIC automatically takes over operation if the primary
NIC fails. Clients maintain their connection with their primary node and,
without disruption, continue to have access to their applications.
Compaq offers a dual-port NIC that can utilize the Compaq Redundant NIC
Utility. This also reduces the possibility of the failure scenario described
above. However, if the entire NIC or the node slot into which the NIC is
placed fails, the same failure scenario will occur.
Compaq Insight Manager and Compaq Insight Manager XE monitor the health
of any network cards used for the LAN. If any of the cards experience a fault,
the Compaq Insight Manager tools mark the card as "Offline" and change its
condition to the appropriate status.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-17
Recommended Cluster Communication Strategy
The past two sections discussed the redundancy of intracluster and
cluster-to-LAN communication. However, to obtain the most benefit while
minimizing cost and complexity, view cluster communications as a single
entity.
To create redundancy for both intracluster and cluster-to-LAN
communication, first, employ physical hardware redundancy for the LAN
NICs. Second, configure the Microsoft clustering software to use both the
primary and redundant LAN NIC as backup for intracluster communication.
With this strategy, your cluster can continue normal operations (without a
failover event) when each of the following points of failure are encountered:
I Failure of the interconnect card
I Failure of the interconnect cable
I Failure of the port on the LAN NIC
I Failure of the LAN NIC (if redundant NICs, as opposed to dual-ported
NICs, are used)
I Failure of the Ethernet cable running from a cluster node to the Ethernet
hub (which connects to the LAN)
The following examples describe how to physically set up your cluster nodes
to employ the Compaq-recommended strategy.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-18 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Example 1
A Compaq dual-port NIC and a single-port NIC are used in this example
(Figure 2-9). The first port of the dual-port NIC is a dedicated interconnect,
and the second port is the backup path for the cluster-to-LAN network. The
single-port NIC is configured as the primary network path for cluster-to-LAN
communication.
The TLAN Teaming and Configuration Utility (for ThunderLAN NICs) and
the Network Teaming and Configuration Utility (for Intel NICs) are used to
configure the second port on the dual-port NIC as the backup port of a
redundant pair. The single port on the other NIC is configured to be the
primary port for cluster-to-LAN communication.
The interconnect retains its fully redundant status when MSCS is configured to
use the other network ports as interconnect backup. Failure of the primary
interconnect path results in intracluster communications occurring over the
single-port NIC, since the single-port NIC was configured in MSCS as the
backup for intracluster communication. If the entire dual-port NIC fails, the
cluster nodes still have a working communication path over the single-port
NIC.
With this configuration, even a failure of the dual-port NIC results in the
transfer of the cluster-to-LAN communication to the single-port NIC. Other
than a failure of the network hub, the failure of any cluster network component
will be resolved by the redundancy of this configuration.
Primary Interconnect Path
Node 2
Node 1
Backup Cluster to LAN and
Primary Cluster to LAN and
Backup Interconnect Path
Backup Interconnect Path
Hub
Clients
Figure 2-9. Use of dual-port NICs to increase redundancy
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-19
Example 2
The second example configuration consists of three single-port NICs (Figure
2-10). One NIC is dedicated to intracluster communication. The other two
NICs are used for cluster-to-LAN communication. The Compaq Advanced
Network Control Utility is used to configure two of the NICs--one as the
primary and one as the standby of a redundant pair.
The interconnect is fully redundant when the Microsoft clustering software is
configured to use the other network cards as backups for the interconnect.
Failure of the primary interconnect path results in intracluster communications
occurring over the primary NIC of the redundant pair. If the entire
interconnect card fails, the cluster nodes will still have a working
communication path.
The cluster-to-LAN communication is fully redundant up to the network hub.
With this configuration, even a failure of the primary NIC results only in the
transfer of the network path to the standby NIC. Other than a failure of the
network hub, any failure of any cluster network component will be resolved by
the redundancy of this configuration.
The primary disadvantage of this configuration as compared to Example 1 is
that an additional card slot is used by the third NIC.
Primary Interconnect Path
Node 1 Node 2
Primary Cluster to LAN and
Backup Interconnect Path
Backup Cluster to LAN and
Backup Interconnect Path
Hub
Clients
Figure 2-10. Use of three NICs to increase redundancy
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-20 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
HA/F100 Fibre Channel Data Paths
The Compaq StorageWorks RAID Array 4000 or Compaq StorageWorks
RAID Array 4100 storage system is the mechanism with which ProLiant
Clusters implement shared storage. Generally, the storage system consists of a
host bus adapter in each server, a storage hub or switch, a Compaq
StorageWorks RA4000 Controller, and a Compaq StorageWorks RAID Array
4000 or Compaq StorageWorks RAID Array 4100 (RA4000/4100) into which
the SCSI disks are placed.
The RA4000/4100 storage system has two distinct data paths, separated by the
Fibre Channel storage hub or FC-AL switch:
I The first data path runs from the host bus adapters in the servers to the
Fibre Channel storage hub or FC-AL switch.
I The second data path runs from the Fibre Channel storage hub or
FC-AL switch to the RA4000/4100.
The effects of a failure will vary depending on whether the failure occurred on
the first or second data path.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-21
Failure of the Host Bus Adapter-to-Storage Hub
Data Path
If the host bus adapter-to-storage hub path fails (Figure 2-11), it results in a
failover of all applications. For instance, if one server can no longer access the
storage hub (and by extension the shared storage), all of the cluster groups that
depend on shared storage will fail over to the second server. The cost of failure
is relatively minor. It is the downtime experienced by users while the failover
event occurs.
RA4000/4100
storage hub
or switch
Interconnect
ProLiant ProLiant
Server Server
Corporate LAN
Figure 2-11. Host bus adapter-to-storage hub data path
Note that the Compaq Insight Manager tools monitor the health of the
RA4000/4100 storage system. If any part of the Fibre Channel data path
disrupts a server's access to the RA4000/4100, the array controller status
changes to "Failed" and the condition is red. The red condition bubbles up to
higher-level Compaq Insight Manager screens and eventually to the device
list.
NOTE: The Compaq Insight Manager tools display a failure of physical hardware through
the Mass Storage button on the View screen, marking the hardware "Failed." A logical
drive in the cluster is reported on the Cluster Shared Resources screen as a logical disk
resource. Compaq Insight Manager and Compaq Insight Manager XE do not associate the
logical drive with the physical hardware.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-22 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Failure of the Hub-to-RA4000/4100 Data Path
The second data path (Figure 2-12), from the storage hub to the RA4000/4100,
has more severe implications when it fails. If this data path fails, all clustered
applications become inoperable. Even attempting to fail the applications to
another cluster node will not gain access to the RA4000/4100.
NOTE: This failure scenario can be avoided by deploying the redundant Fibre Channel
loop configuration of the Compaq ProLiant Cluster HA/F200.
RA4000/4100
storage hub
or switch
Interconnect
ProLiant ProLiant
Server Server
Corporate LAN
Figure 2-12. Hub-to-RA4000/4100 data path
Without access to shared storage, clustered applications cannot reach their data
or log files. The data, however, is unharmed and remains safely stored on the
physical disks inside the RA4000/4100. If a database application was running
when this failure occurred, some in-progress transactions will be lost. The
database will need to be rolled back and the in-progress transactions re-
entered.
Like the server-to-storage hub data path, the Compaq Insight Manager tools
detect this fault, change the RA4000/4100 status to "Failed," and change its
condition to red. The red condition bubbles up through Compaq Insight
Manager screens, eventually to the device list.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-23
Nonshared Disk Drives
Nonshared disk drives, or local storage, operate the same way in a cluster as
they do in a single-server environment. These drives can be in the server drive
bays or in an external storage cabinet. As long as they are not accessible by
both servers, they are considered nonshared.
Treat nonshared drives in a clustered environment as you would in a
nonclustered environment. Most likely, some form of RAID is used to protect
the drives and restore a failed drive. Since the operating system is stored on
these drives, use either hardware or software RAID to protect the information.
Hardware RAID is available with the Compaq SMART-2 Controller or by
using a nonshared storage system.
Shared Disk Drives
Shared disk drives are contained in the RA4000/4100, which is accessible by
each cluster node. Employ hardware RAID 1 or 5 on all of your shared disk
drives. This is configured using the Compaq Array Configuration Utility.
If RAID 1 or 5 is not used, failure of a shared disk drive will disrupt service to
all clustered applications and services that depend on the drive. Failover of a
cluster node will not resolve this failure, since neither server can read from a
failed drive.
NOTE: Windows NTS/E software RAID is not available for shared drives when using
MSCS. Hardware RAID is the only available RAID option for shared storage.
As with other system failures, Compaq Insight Manager monitors the health of
disk drives and will mark a failed drive as "Failed."
Enhanced High Availability Features of the HA/F200
A single point of failure refers to any component in the system that, should it
fail, prevents the system from functioning. Single points of failure in hardware
can be minimized, and in some cases eliminated, by using redundant
components. The most effective way of accomplishing this is by clustering.
The Compaq ProLiant Cluster HA/F100 reduces the single points of failure
that exist in a single-server environment by allowing two servers to share
storage and take over for each other in the event that one server fails. The
Compaq ProLiant Cluster HA/F200 goes one step further by implementing a
dual redundant Fibre Channel Arbitrated Loop configuration.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-24 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
The Compaq ProLiant Cluster HA/F200 further enhances high availability
through the use of additional, redundant, components in the server-to-storage
connection and in the shared storage system itself. In the event of a failure,
processing is switched to an alternate path without affecting applications and
end users. In fact, this path switch is transparent even to the Windows NT and
Windows 2000 file system (NTFS). The combination of multiple paths and
redundant hardware components provided by the HA/F200 offers significantly
enhanced high availability over non-redundant configurations.
A single component failure in the HA/F200 will result in an automatic failover
to an alternate component, allowing end users to continue accessing their
applications without interruption. Some typical failures and associated
responses in an HA/F200 configuration are:
I A server failure will cause the Microsoft clustering software to fail
application processing over to the second server.
I A host bus adapter failure will cause I/O requests intended for the failed
adapter to be rerouted through the remaining adapter.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-25
I A storage hub, switch, or cable failure will be treated like a host bus
adapter failure and a failover to the second host bus adapter, which is
using a different storage hub and cables, will occur.
I An array controller failure will cause the redundant array controller to
take over for the failed controller.
In all of the above examples, end users will experience minimal interruptions
while the failover occurs. In some cases, the interruptions may not even be
noticeable.
The following illustration depicts the HA/F200 configuration components.
Node 1
RA4000/4100
storage hub
or switch
Dedicated
Interconnect
storage hub
or switch
Node 2
LAN
Figure 2-13. HA/F200 configuration
HA/F200 Fibre Channel Data Paths
The Compaq StorageWorks RAID Array 4000/4100 storage system is the
mechanism with which the HA/F200 cluster implements shared storage. The
Compaq ProLiant Cluster HA/F200 minimum configuration consists of two
host bus adapters in each server, two Fibre Channel storage hubs or FC-AL
switches, two array controllers per RA4000/4100, and one or more
RA4000/4100s.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-26 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
The RA4000/4100 storage system has active data paths and standby data
paths, separated by two Fibre Channel storage hubs or FC-AL switches. Figure
2-14 and Figure 2-15 detail the active and standby paths of the minimum
HA/F200 configuration.
A A
S S
Server Server
storage hub storage hub
or switch or switch
Active Standby
RA4000/4100
Figure 2-14. Active host bus adapter-to-storage data paths
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-27
The active data paths run from the active host bus adapters in the servers to the
active storage hub. If this path fails, the applications can seamlessly fail over
to the standby host bus adapter-to-storage hub data paths (Figure 2-15).
A A
S S
Server Server
storage hub storage hub
or switch or switch
Active Standby
RA4000/4100
Figure 2-15. Active hub-to-storage data path
The second active data path runs from the active hub or switch to the
RA4000/4100. If this path fails, the applications can seamlessly fail over to the
standby hub-to-RA4000/4100 data path.
The dual redundant loop feature of the Compaq ProLiant Cluster HA/F200
increases the level of availability over clusters that have only one path to the
shared storage. In addition, the second path in the HA/F200 provides for
improved performance through static load balancing. Static load balancing
considerations are discussed in the "Static Load Balancing" section of this
chapter.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-28 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Capacity Planning
Capacity planning determines how much computer hardware is needed to
support the applications and data on your clustered servers. Unlike
conventional, single-server capacity planning, clustered configurations must
ensure that each node is capable of running any applications or services that
may fail over from its partner node. To simplify the following discussion, the
software running on each of the clustered nodes is divided into three generic
categories:
I Operating system
I Nonclustered applications and services
I Clustered applications and services
Figure 2-16 illustrates these categories in the cluster.
Data for Node1 Clustered
Applications & Services
Data for Node2 Clustered
Applications & Services
Shared Storage
Operating System
Operating System
Clustered Applications
Clustered Applications
& Services
& Services
Non-Clustered
Non-Clustered
Applications & Services
Applications & Services
Node2
Node1
Figure 2-16. File locations in a Compaq ProLiant Cluster
For each server, determine the processor, memory, and disk storage
requirements needed to support its operating system and nonclustered
applications and services.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-29
Determine the processor and memory requirements needed to support the
clustered applications and services that will run on each node while the cluster
is in a normal operating state.
If the program files of a clustered application and/or service will reside on
local storage, remember to add that capacity to the amount of local storage
needed on each node.
For all files that will reside on shared storage, see "Shared Storage Capacity"
later in this chapter.
Server Capacity
The capacity needed in each server depends on whether you design your
cluster as an active/active configuration or as an active/standby configuration.
Capacity planning for each configuration is discussed in the following
sections.
Active/Active Configuration
As described earlier in this chapter, an active/active configuration can be
designed in two ways:
I Applications and services may be configured to fail over from each node
to its partner node.
I Applications and services may be configured to fail over from just one
node to its partner node.
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-30 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
The following table details the capacity requirements that can be applied to
either active/active design.
Table 2-1
Server Capacity* Requirements for Active/Active Configuration
Node 1 Node 2
Operating system (with MSCS) Operating system (with MSCS)
Nonclustered applications and services Nonclustered applications and services
Server1 clustered applications and services Server2 clustered applications and services
Server2 clustered applications and services Server1 clustered applications and services
(if Server2 is set up to fail applications and (if Server1 is set up to fail applications and
services to Server1) services to Server2)
* Processing power, memory, and nonshared storage
Active/Standby Configuration
In an active/standby configuration, only one node actively runs applications
and services. The other node is in an idle, or standby, state. Assume Node 1 is
the active node and Node 2 is the standby node.
Table 2-2
Server Capacity* Requirements for Active/Standby Configuration
Node 1 Node 2
Operating System (with MSCS) Operating system (with MSCS)
Nonclustered applications and services Server1 clustered applications and
services
Server1 clustered applications and services
* Processing power, memory, and nonshared storage
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
Designing the Compaq ProLiant Clusters HA/F100 and HA/F200 2-31
Shared Storage Capacity
Each server is connected to shared storage (the Compaq StorageWorks RAID
Array 4000/4100 storage system), which mainly stores data files of clustered
applications and services. Follow the guidelines below to determine how much
capacity is needed for your shared storage.
NOTE: For some clustered applications, it may make sense to store the application
program files on shared storage. If the application allows customization and the
customized information is stored in program files, the program files should be placed on
shared storage. When a failover event occurs, the secondary node will launch the
application from shared storage. The application will execute with the same
customizations that existed when executed on the primary node.
Two factors help to determine the required amount of shared storage disk
space:
I The amount of space required for all clustered applications and their
dependencies.
I The level of data protection (RAID) required for each type of data used
by each clustered application. Two factors driving RAID requirements
are:
The performance required for each drive volume
The recovery time required for each drive volume
IMPORTANT: Windows software RAID is not available for shared drives when using
MSCS. Hardware RAID is the only available RAID option for shared storage.
For more information about hardware RAID, see the following:
I Compaq StorageWorks Fibre Channel RAID Array 4000 User Guide
I Compaq StorageWorks Fibre Channel RAID Array 4100 User Guide
Compaq Confidential Need to Know Required
Writer: Bryan Hicks Project: Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide Comments:
Part Number: 380362-003 File Name: c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc Last Saved On: 8/24/00 12:00 PM
2-32 Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
In the "Cluster Groups" section of this chapter, you created a resource
dependency tree, then transferred that information into a Cluster Group
Definition Worksheet (Figure 2-8). Under the resource dependencies in the
worksheet, you listed at least one physical disk resource. For each physical
disk resource, determine the capacity and level of protection required for the
data to be stored on it.
For example, the Web Sales Order Database group depends on a log file, data
files, and program files. It might be important for the log file and program files
to have a quick recovery time, while performance would be a secondary
concern. Together, the files do not take up much capacity; therefore, mirroring
| 380362-003
|
