ActiveAnswers
iss: business applications solutions
january 2003
technical guide
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
abstract: PeopleSoft 8 High Availability solutions currently require the use of four ProLiant servers for optimum redundancy and fault tolerance across the Database, Application, and Web servers. This guide intends to provide an overview of the implementation of PeopleSoft 8 with a loadbalanced Web server farm in a Microsoft-certified cluster environment. It also provides a section with answers and explanations of questions associated with setting up such an environment.
This guide provides you with the basic building blocks of information to assist in the setup and configuration of both Microsoft SQL Server and the PeopleSoft 8 application in a clustered Microsoft Windows 2000 Advanced Server environment. It is a supplemental document intended to be used with specific knowledge and/or the appropriate installation and configuration guides for PeopleSoft, Microsoft SQL Server and Microsoft cluster service.
contents
introduction..................... 3 assumptions................ 3 overview ......................... 4 HP-PeopleSoft cluster solutions.......................... 5 software requirements............... 6 hardware requirements............... 6 HP/PeopleSoft cluster solution hardware configurations ............. 8 HP/PeopleSoft cluster solution – software configurations ........... 14 Conclusion .................... 17 Appendix A: Frequently Asked Questions...................... 19
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
2
notice
The information in this document is subject to change without notice. Microsoft, Windows and Windows NT are either registered trademarks or trademarks of Microsoft Corporation in the United States and/or other countries. Intel, Pentium and Pentium® III Xeon are trademarks and/or registered trademarks of Intel Corporation. january 2003 P/N 5981-4961ena1
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
3
introduction
The environment discussed in this guide consists of the following: two servers configured in a clustered environment running Microsoft Cluster Server for the application and database servers, and two servers configured as Web servers load-balanced in the web server farm. The application server and the database server represent the two members of the cluster, which are able to failover to each other if needed. In the event of a database/application tier failure, a physical 4-tier environment running the database and application tier on two servers becomes a logical 4-tier environment running on one server. The implementation of a Fibre storage system environment for the shared storage of the cluster is recommended. The Web servers operate in a Distributed Internet Server Array (DISA) architecture on a separate tier. The Web servers are configured in a Network Load Balancing (NLB) cluster or using a hardware-based load-balancing appliance solution for fault tolerance. Multiple hardware-based load-balancing appliances can be deployed to eliminate single points of failure in the Web server farm. In the event of a Web tier server failure, the loadbalancing component removes the failed server from the configuration and balances the traffic over the remaining surviving nodes. For more details on these various solutions areas, refer to the ProLiant Servers and StorageWorks Storage Area Network (SAN) in a PeopleSoft 8 Environment Solutions Guide and the Overview of the HP Dynamic Internet Solutions Architecture (DISA) document at the website for ActiveAnswers .
assumptions
documentation
Since this is an implementation overview, users of this guide should have access to the latest version of the following documentation: • • • ProLiant Servers and StorageWorks Storage Area Network (SAN) in a PeopleSoft 8 Environment Microsoft Step-by-Step Guide to Installing Cluster Service PeopleSoft 8 Installation and Configuration (from PeopleSoft)
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
4
• •
Microsoft Configuring Network Load Balancing Or Intel NetStructure 7280 XML Director (or equivalent provided with the product)
knowledge
This guide assumes sufficient technical knowledge of the following systems, applications, and procedures: • • • • • • • Microsoft SQL Server Configuration, Implementation and Administration Microsoft Cluster Service Installation and Configuration PeopleSoft 8.12.10 Installation and Configuration Microsoft Windows 2000 Advanced Server General Windows 2000 Networking Knowledge Microsoft Network Load Balancing (NLB) Configuration Or Intel NetStructure 7280 XML Director (or equivalent) Configuration
overview
Even though server hardware has become more reliable through the use of redundant and hot-pluggable components, it is still far from providing “100% availability”. It is common nowadays to hear of requirements of “99.5% availability” or “99.9% availability” for large PeopleSoft multi-tier enterprise server environments. What does this really mean in terms of “How much server downtime does it represent per year?” and the terminology used to achieve the targeted uptime.
Note: HP Services is offering 99.5% as well as 99.9% packaged Uptime Guarantees for Windows NT/2000 on ProLiant servers. See the HP High Availability Uptime Guarantee website for the exact terms and conditions of these offers. Also, you can contact HP Global Services for a custom HP Global High Availability service offering at http://h18005.www1.hp.com/services/available/
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
5
Table 1. High Availability Uptime
% Uptime 99 99.5 99.9 99.99 99.999 99.9999 Max downtime/year 87 Hours 43 Hours 8.5 Hours 1 Hour 5 Minutes 32 Seconds Technology Conventional High Availability High Availability Fault Resilience Fault Tolerance Fault Tolerance
As can be seen from Table 1, providing a 99.9% uptime to end-users allows for eight hours of downtime in a year. This includes planned (maintenance) and unplanned outages (failure). This downtime does not take into consideration the many outages due to operator error. The issue of system availability motivated the introduction of the serverclustering paradigm. The server-clustering paradigm presents the following features: • High Availability –to ensure as little downtime as possible by providing failing over of resources between nodes (also known as a resource failover) in case of node failure or simply upon management request Combined power –to provide computing power scalability in the cluster by balancing a given load on multiple nodes also known as load balancing Ease of management – operations in a cluster are carried out only once for all members of that cluster, no matter how many nodes are present (using a cluster is no different from using a single node as it appears to system managers and to end-users as a Single System Image) also known as a Virtual Server on Windows 2000
• •
Through the use of clusters, two ProLiant servers running Microsoft Windows 2000 Advanced Server can be configured to provide maximum availability, combined power, and ease of management.
HP-PeopleSoft cluster solutions
Today’s clustering technology is designed to keep server-based applications highly available, regardless of individual component failures. This section provides the following information: • The basic software and hardware requirements to configure a HP-PeopleSoft cluster solution
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
6
• •
The logical flow of an HTML-based client transaction for a PeopleSoft 8 high availability clustered environment A description of the PeopleSoft 8 application cluster/server components along with illustrations of the system modes in the event of an application, database, or Web server failure.
software requirements
The software configurations discussed in this guide are: • Application: PeopleSoft 8 Applications PeopleTools 8.12.10 PeopleSoft Edition of BEA TUXEDO v6.5/BEA Jolt 1.1 PeopleSoft Internet Architecture 8.12.10 BEA WebLogic 5.1.0 • • Database: Cluster: Microsoft SQL Server 2000 Enterprise Edition Microsoft Cluster Service (incl. with Windows 2000 Advanced Server) Microsoft Network Load Balancing (incl. with Windows 2000 Advanced Server as a service) • • OS: Support: Microsoft Windows 2000 Advanced Server (Build 2195) SmartStart 5.1 or greater Management CD 5.1 or greater Support Paq (CSP) for Windows 2000 (from SmartStart CD) Array Configuration Utility (ACU) (from SmartStart CD) Survey Utility (from Management CD) Insight Manager Server Agents (from Management CD)
hardware requirements
To avoid hardware incompatibilities, HP recommends that you use a hardware configuration recommended by the HP PeopleSoft Competency Centers. They can be contacted by e-mail mailto:at mcset_tco@hp.com
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
7
A sample configuration recommended for the application and database server nodes are as follows: • Two ProLiant DL760 servers, similarly configured to include: – – – Dual Port NIC (Netelligent Dual 10/100TX PCI UTP Controller) Two 4GB (minimum) logical disk partitions Enough disks to support an external database. A private network interconnect used for cluster heartbeats and other cluster protocol exchanges (Null 100Mbit Network Crossover Cable can be used) An interconnected Fibre Channel disk subsystem. Can be located on a Storage Area Network (SAN) such as the StorageWorks MSA1000 SAN or the StorageWorks Modular Array 8000 / Enterprise Modular Array 12000 (MA8000/EMA12000) or dedicated storage for the cluster (that is, MSA1000 connected via a fibre channel switch/hub).
–
A sample recommended configuration for the Web server nodes is as follows: • Two ProLiant DL360 servers, similarly configured to include: – – Dual Port Embedded NIC (Netelligent Dual 10/100TX PCI UTP Controller) Two 4GB (minimum) logical disk partitions
HP offers many additional cluster configurations in its support of high availability. These solutions can be complex and may contain many components or even a cluster package. The main components of a cluster are: • • • Server −Most models of the ProLiant server platforms are supported in a cluster. Interconnect −HP strongly recommends the use of a private network connection between cluster nodes. Storage – Most external StorageWorks RAID Array storage systems are supported. The maximum number of storage subsystems that can be attached to a server is dependent on the server. HP supports both SCSI and fibre optical external storage systems connections. Operating System – Both Windows NT 4.0 Server Enterprise Edition and Windows 2000 Advanced Server are supported. Cluster Kits – HP offer cluster kits that include cables, documentation, and clustering software utilities.
• •
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
8
The website http://h18000.www1.hp.com/solutions/enterprise/highavailability/index.html lists the latest supported servers and server combinations, interconnects, storage subsystems and cluster kits that are necessary for cluster implementations. Documentation of clustering configurations can also be obtained from this website.
HP/PeopleSoft cluster solution - hardware configurations
Figure 1 illustrates the initial hardware configuration of the clustered database and application servers with the load-balanced Web servers.
Figure 1 – Initial PeopleSoft 8 Hardware Layout
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
9
Figure 2 illustrates the logical flow of HTML-based client transactions. PeopleSoft application and Microsoft SQL Server are installed in a standard running mode in a Microsoft cluster. PeopleSoft is running in a physical 4-tier environment with the database on one server and the application residing on the other server. The Web servers are loadbalanced with a hardware-based load balancer.
Client connects to Director across the network topology Director selects server to fulfill client request
A B 1.Do I have specific rules for request? Port/URL No No 2.Is server working? Yes Yes 3.How fast is server responding? 50ms 35ms Client requests are switched to the fastest server – automatically. This selection process is completely dynamic and adjusts to server loads automatically BEA TUXEDO listens for client request from Web Server farm Client requests are directed to the active PeopleSoft Services domain within the cluster by Cluster Services PeopleSoft application logic processes the request and retrieves the data requested from the active database node within the cluster
Figure 2: PeopleSoft 8 HTML-based Client Transaction Logical Flow
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
10
As shown in Figure 3, PeopleSoft is installed in a standard running mode in a Microsoft cluster. PeopleSoft is running in a physical 4-tier environment with the database on one server and the application residing on the other server.
Figure 3: Initial Software Configuration
The system in Database Server Failure Mode is illustrated in Figure 4. PeopleSoft is running in a logical 4-tier environment with the database server and the application server running on the same system.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
11
Figure 4: Database Server Failure
Figure 5 illustrates the system in the Application Server Failure Mode. PeopleSoft is running in a logical 4-tier environment with the database server and the application server both running on the same system.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
12
Figure 5: Application Server Failure
As illustrated in Figure 6, the system is in Web Server Failure Mode while participating in a Network Load Balancing (NLB) cluster. The Web servers are configured with multiple network adapters in unicast mode. PeopleSoft continues to run in a physical 4-tier environment with the surviving Web server handling the Web client load. The database server and the application server both continue to run on separate systems.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
13
Figure 6: Web Server Failure Running Network Load Balancing (NLB) As illustrated in Figure 7, the system is in Web Server Failure and/or Hardware Load Balancer Failure Mode. The Web servers are configured with multiple network adapters and redundant Load Balancers are deployed to eliminate single points of failure in the server farm. PeopleSoft continues to run in a physical 4-tier environment with the surviving Web server and Load Balancer handling the Web client load. The database server and the application server both continue to run on separate systems.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
14
Figure 7: Web Server and/or Hardware Load Balancer Failure
HP/PeopleSoft cluster solution – software configurations
The system in a standard running mode is illustrated in Figure 8. PeopleSoft is running in a physical 4-tier environment with the database on one server and the application residing on the other server.
Note: PeopleSoft Edition of BEA TUXEDO v6.5/BEA Jolt 1.1 runs on both systems at the same time; however, only one is active.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
15
Figure 8: Initial Software Configuration
As illustrated in Figure 9, the Cluster Group containing the PeopleSoft application service migrates to the database server if the application server fails.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
16
Figure 9: Configuration after Application Server Failure
As illustrated in Figure 10, the Cluster Group containing the PeopleSoft database resources migrates to the application server if the database server fails.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
17
Figure 10: Configuration after Database Server Failure
As can be seen from the illustrations, understanding the PeopleSoft application program architecture is a key requirement for implementing high availability. The PeopleSoft application is designed around: the client/server architecture, the IP network protocol, and the ability of a client to communicate with its server. The server components are candidates for cluster failovers as they are designed to maintain data integrity. The client components can disconnect from the failed server component and re-establish the connection with the surviving server component. Most PeopleSoft application components can be maintained at a highly available level by allowing one server component to switch to an alternative when failure occurs. The client software is designed to experience little or no interruption when the function moves from one server to the other.
Conclusion
The ability of a server or an application solution to tolerate hardware and software failures is a key requirement in implementing high availability. High availability usually implies system uptime of 99.5 - 99.9% and higher. The use of an alternative hardware source composed of redundant software using the same data path is one way in which high availability can be implemented with today’s cluster technology. Clustering protects the availability of the data, but not the data itself.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
18
HP and PeopleSoft have developed failover configurations that test the behavior of the PeopleSoft 8 application while making the database and application servers highly available. These configurations were tested for application server, Web server, and database server failures. Tests were successfully conducted to provide general guidelines to configure high availability servers as detailed in this document. The manner in which the PeopleSoft 8 application tolerates its software failures is a key requirement for implementing high availability in a PeopleSoft 8 application environment. PeopleSoft 8 is designed around the client/server architecture, the IP network protocol, and the ability of a client to communicate with its server. The server components are candidates for cluster failovers as they are designed to maintain data integrity. The client component is able to disconnect from the failed server component and re-establish connection with a new server component. HP offers many high availability configurations that increase a server’s availability by allowing a fully automated switchover of server operations to another server. These configurations support Microsoft Windows 2000 Advanced Server operating system as well as a majority of disk controllers and StorageWorks subsystems. They can be used to make PeopleSoft 8 application server components highly available.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
19
Appendix A: Frequently Asked Questions
This section details the most frequently asked questions about the Microsoft, PeopleSoft, and Microsoft SQL Server cluster solution provided by HP. This information along with the appropriate installation and configuration guides will assist in the set up of the PeopleSoft 8 clustered environment. 1. Which versions of SQL Server and PeopleSoft software were tested? The following versions of SQL Server and PeopleSoft Software were tested in cooperation with the Alliance Solution Center. – – – – – Microsoft SQL Server 2000 Enterprise Edition PeopleSoft 8.12.10 PeopleTools PeopleSoft 8 Applications PeopleSoft Edition of BEA TUXEDO v6.5/BEA Jolt 1.1 PeopleSoft Internet Architecture 8.12.10
PeopleSoft currently does not certify any hardware vendor's high availability software. However, PeopleSoft and HP have successfully completed projects testing this Microsoft Cluster Service Solution with PeopleSoft applications. 2. Is the Microsoft Cluster Service failover automatic or manual? It is completely automatic for both hardware failure and software failure. In addition, the recovery after a failure correction is also completely automatic with no user intervention required. 3. How long does it take the database and/or application server to be operational again when failover occurs? It will vary from system to system depending on the size of the database, the number of users working on the system at the time of failure, and the amount of time since the last checkpoint. Typical recovery time is typically from 20 seconds to 3 minutes.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
20
4. Are there any special steps required to implement PeopleSoft failover? Yes, but they are not complicated. Contact the HP PeopleSoft Competency Centers for details. Note: The HP PeopleSoft Competency Centers can be contacted by e-mail at: mailto:mcset_tco@hp.com
5. How many nodes does Microsoft Cluster Service support? Currently, Microsoft with the Windows 2000 Advanced Server Cluster Service supports a maximum of two cluster nodes. A two-node cluster will be an ideal solution for a single instance database. You can configure multiple instances in a two-node cluster configuration, but nodes in the cluster should be sized properly keeping in mind that in case of failover, all the instances will be on one system. 6. How many nodes does Microsoft support in a Network Load Balancing (NLB) Cluster? Currently Microsoft with the Windows 2000 Advanced Server Network Load Balancing Service supports a maximum of 32 nodes. A two-node NLB cluster is the minimum solution for an implementation however the cluster should be sized properly keeping in mind that in case of a Web server failure all the Web traffic will be distributed among the surviving Web servers. Please contact the HP PeopleSoft Competency Centers for an appropriate sizing. 7. Do you need Microsoft Cluster Service failover for the PeopleSoft application server? In general, the PeopleSoft application server has built-in High Availability (HA) features so it is not required to have Microsoft Cluster Service on the application server. Multiple application servers would be used for built-in load balancing and redundancy. However, in this four- server configuration the application server and the database server are configured for failover to each other using the cluster service and therefore it is required for fault tolerance. 8. What is the impact on users that are logged-in but idle? There is no impact on idle users. Unless a transaction is issued during the failure window, they will see nothing.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
21
9. How are user transactions affected by failover? It depends on the phase of the transaction process. In most cases, a failover during a transaction causes an error at the client that requires a query to be resubmitted. This is completely dependent on the amount of time allocated to the client query timeout and the amount of time it takes for the PeopleSoft environment to come back online. If the failover occurs before or after a transaction, there is no impact, the transaction is successful. In some cases, even during a transaction, a failover can cause the application server to resubmit the transaction and it succeeds with no error to the client. 10. What is the impact on users who have submitted a transaction but it has not been committed? See the answer to 9. 11. What is the impact on users who have submitted a transaction that has been committed but have not yet received confirmation? See the answer to 9. 12. Are there any deviations from the standard PeopleSoft installation procedure for a High Availability environment? Actual installation of PeopleSoft software in a high availability configuration remains the same. There are however, steps required to cluster the database, and then the step of clustering the PeopleSoft application service. All this is done and controlled using the Microsoft Cluster Service and PeopleSoft software. Additionally, there are steps required to either cluster the Web servers with the Network Load Balancing (NLB) service or configure a hardware-based load balancer solution. 13. Are there alternative cluster hardware configurations?
The following are other possible configurations. However, they are not usually recommended for the reasons given in each scenario. • Active/Active Database Only - This configuration utilizes two database servers, each running their own database that can failover to the other server, and four application servers. This configuration requires that BEA Tuxedo failover features be used to eliminate the single point of failure for the application servers.
5981-4961ena1
implementing PeopleSoft 8 applications in a Microsoft clustered environment on ProLiant servers
22
The advantages of this configuration are: – – – Both database servers in the cluster can be configured and tuned for optimal database performance The resources on each database server are being utilized in parallel and are operational under normal conditions The application servers are configured strictly and optimally for business logic use.
The disadvantage is that in the event of a database server failure, the combined load of both databases residing on the single remaining database server would be extremely high and overall system performance may be affected. • Active/Passive Database Only - This configuration utilizes two database servers, one running its own database and the other one idling or performing minimal tasks while on standby (should the other database server fail). This configuration requires that two application servers be utilized and that BEA Tuxedo failover features be used to eliminate the single point of failure for the application servers.
The advantages of this configuration are: – – Both database servers in the cluster can be configured and tuned for optimal database performance In the event of a database server failure, overall system performance should not be adversely affected when both database server nodes have been configured identically The application servers are configured strictly and optimally for business logic use.
–
The disadvantage is the capital expenditure on a database server that only idles may not be considered cost effective.
5981-4961ena1