step-by-step installation of platform suite for sas on...

Step-by-Step Installation of

Platform Suite for SAS on Windows A visual guide to getting your grid up and running.

05Jan2007 of 41

Table of Contents

INTRODUCTION ........................................................................................................................................ 3

OVERVIEW.................................................................................................................................................. 3 NOTES........................................................................................................................................................ 4

INSTALLATION ......................................................................................................................................... 5

PRE-INSTALLATION TASKS......................................................................................................................... 5 INSTALLING PROCESS MANAGER AND LSF ON THE GRID CONTROL MACHINE (LSF MASTER) ................. 5 TESTING THE GRID CONTROL MACHINE INSTALLATION .......................................................................... 18 INSTALLING LSF ON GRID NODE MACHINES & FOUNDATION SAS CLIENTS ........................................... 19 TESTING THE GRID NODE INSTALLATION ................................................................................................. 32 ADDING GRID NODES OR FOUNDATION SAS CLIENTS TO THE CLUSTER.................................................. 33 INSTALLING GRID MANAGEMENT SERVICE (GMS).................................................................................. 34 POST INSTALLATION TASKS ..................................................................................................................... 39

CONVERTING A GRID NODE MACHINE TO A GRID CLIENT .................................................... 40

LSF QUICK REFERENCE....................................................................................................................... 41

05Jan2007 of 41

Introduction

Overview

Several types of machines make up a SAS grid environment. These machines have been defined to clarify the software components that must be installed on each one as well as the SAS metadata that must be configured. The SAS metadata server is shown on a separate machine in this sample architecture. It is often common to dedicate a machine to running the SAS metadata server, but you may choose to run the metadata server on the grid control machine. The SAS metadata server installation and configuration should be the very first step, as is typical in a SAS 9 multi-machine install. The three machines specific to a grid installation are defined as follows:

• grid client - a grid client submits work to the grid but is not part of the grid resources available to execute work. Examples of a grid client would be:

o a SAS Data Integration Studio client (Platform LSF not installed on this client machine)

o a SAS Enterprise Miner client (Platform LSF not installed on this client machine)

o a SASMC client using the Schedule Manager plug-in or any other applications scheduling SAS workflows. (Platform LSF not installed on this client machine)

o a SAS foundation install (minimum Base SAS, SAS/CONNECT and Platform LSF) used to run a program that submits work, both whole programs or programs broken into parallel chunks, to the grid. Installation of the Platform LSF component is required in this case in order for SAS/CONNECT to submit the work to the grid.

05Jan2007 of 41

• grid control machine - any machine in the grid can be designated as the grid control machine. More software is installed on the grid control machine and more SAS metadata configuration takes place on this machine. You should start the installation of the Platform Suite for SAS on this machine. In a SAS Data Integration Studio and SAS Enterprise Miner scenario the grid control machine runs a workspace server that executes programs that utilize SAS/CONNECT to distribute work to the grid nodes. The grid control machine can be configured as a grid resource capable of receiving work to execute or not, depending on the needs of your environment.

• grid node - a grid node is a grid computing resource capable of receiving the work that is being distributed. Each grid node must be running a minimum of Base SAS, SAS/CONNECT and Platform LSF.

A SAS/CONNECT client could be run on any machine within the grid and also may be capable of receiving work to execute.

Installation of Platform Suite for SAS is performed first on the Grid Control Machine and is followed by installation on all of the Grid Node machines. Installation of Platform LSF on the Grid Control Machine can be installed as a part of the Process Manager installation or it can be installed by itself. This document will only show LSF being installed as part of the Process Manager installation. Platform LSF is installed using one of two methods, based upon whether the machine is running a Windows based operating system or a Unix based operating system. This document will walk you step by step through each. Not only do machines that do processing for the grid need Platform LSF installed, but machines that submit jobs to run on the grid must also have Platform LSF installed on them. Grid clients such as DI Studio or Enterprise Miner do not submit jobs directly but rather work with a SAS Workspace Server or a Stored-Process Server that does the job submission. Since those grid clients do not submit jobs, they do not need Platform LSF installed, but the machine where the Workspace Server or Stored-Process server would need it installed. If you are writing your own grid-enabled SAS program in Foundation SAS and want to run the program, that grid client workstation must have Platform LSF installed since it will be doing the actual submission of jobs to the grid.

Notes • Although the screenshot requests a <domain>\userid it is recommended that you

refer to the relevant section on User Accounts in the SAS Intelligence Platform: Application Server Administration Guide for advice pertinent to your specific installation needs. This installation assumes it will be ‘<domain>\lsfadmin’.

• This document assumes you have received the Platform Suite for SAS license file (i.e., PlatformJSxxxxxx.txt) either with your software bundle or by ordering it from http://support.sas.com/adminservices-SID.

• This document will assist you installing Platform LSF to create a computer cluster, but to enable the cluster to work with the SAS Business Intelligence Platform, please refer to the Supporting Grid Computing chapter in the SAS Intelligence Platform: Application Server Administrative Guide. This section is available online at http://support.sas.com/onlinedoc/913/docMainpage.jsp?topic=biasag.hlp/a003087200.htm.

05Jan2007 of 41

Installation

Pre-installation Tasks 1. Create a domain LSF administrator account.

NOTE: Although the screenshot requests a <domain>\userid it is recommended that you refer to the relevant section on User Accounts in the SAS Intelligence Platform: Application Server Administration Guide for advice pertinent to your specific installation needs. This installation assumes it will be ‘<domain>\lsfadmin’.

Installing Process Manager and LSF on the Grid Control Machine (LSF Master) 1. Run pm3.0_pinstall_sas_win.exe. It is located on the 5th Platform Suite for SAS CD.

05Jan2007 of 41

2. You should be presented with a window showing the extraction progress.

3. When extraction is complete, the Platform Process Manager screen is displayed.

05Jan2007 of 41

4. When the installation program has completed loading, it will display the Platform Process Manager welcome screen.

Press ‘Next’ to continue.

5. Next, the Platform End User License Agreement is displayed.

Press ‘Agree’ or ‘Disagree’ after reading the agreement (the default is ‘Disagree’).

05Jan2007 of 41

6. Next the install asks you for the location to install the Platform Process Manager files.


7. Next, the install will ask you what components you want to install.

Since we assume you are installing both Process Manager and LSF, make sure everything is selected. If LSF is already installed, remove the check next to the ‘Platform LSF’ option. Press ‘Next’ to continue (the default is ‘Back’).

05Jan2007 of 41

8. Next, the install will ask you the name of the cluster’s administrator and the path of the license file.

You should enter the domain LSF administrator you created in the SAS pre-installation checklist (usually <domain>\lsfadmin where <domain> is your Windows domain name). Find the ‘PlatformJSxxxxxx.txt’ file provided with your SID file and put its path into the dialog. Leave the port number at 1966. Press ‘Next’ to continue.

05Jan2007 of 41

9. Next, the install will ask you for the user ID that will be used to run the Process Manager and LSF service programs.

You should enter the domain LSF administrator you created in the SAS pre-installation checklist (usually <domain>\lsfadmin where <domain> is your Windows domain name).

10. Next, LSF will give you the option of receiving email whenever a job completes.

Leave the box unchecked and press ‘Next’ to continue.

05Jan2007 of 41

11. The install has gathered all the information it needs and will present you with a summary before starting the install.


12. The install will display the progress as it installs Progress Manager.

05Jan2007 of 41

13. Next, LSF needs to know the name of the cluster. This name will be used to identify LSF services and configuration files.

Enter the name you want to call the cluster and press ‘Next’ to continue.

14. Next, LSF needs to know the name of the computer you want to install LSF.

Enter the fully qualified domain name (FQDN) of your grid control machine (i.e., gridcontrol.domain.company.com). Adding the domain qualification to the name will eliminate problems with multi-homed machines and clusters that include operating systems other than Windows.

05Jan2007 of 41

After pressing the ‘->’ button the name should appear in the right hand list box.


15. Next, LSF wants to know where it should install its files.

Enter a new path or take the default and press ‘Next’ to continue.

05Jan2007 of 41

16. Next, LSF needs to share the previously selected directory on the LAN so that any additional computers attaching to the cluster can read configuration files.

This will usually be ‘\\<Grid_Control_FQDN>\<LSF_install_directory>’. You can change the shared directory name, but do not change the hostname. Press ‘Next’ to continue.

LSF will then put a confirmation dialog to make sure you know what it is going to do by creating the share.

05Jan2007 of 41

17. LSF has completed getting all the information it needs and displays it in a summary before starting the install.

Press ‘Next’ to start the install and display the progress indicator.

05Jan2007 of 41

18. When the install is complete, LSF will display an installation results summary.

Make sure everything succeeded and press ‘Next’ to continue.

19. LSF has completed the install and displays its final screen.

Keep the checkbox selected so that LSF will display the “LSF Get Started” file and press ‘Finish’ to continue.

05Jan2007 of 41

20. The “LSF_Get_Started.txt” file is displayed.

As shown, you must add the user account and password of the users that will be submitting jobs to the cluster by running the lspasswd program from a command prompt. This can only be done after you reboot the machine as requested in step 22. Once you have read the file, close Notepad.

21. Next, LSF will tell you that the computer needs to be rebooted.

22. Once the system has rebooted, use lspasswd to add the lsfadmin, sasadm, sastrust,

and sasdemo user ids since those accounts will be used to submit jobs to the cluster.

05Jan2007 of 41

Testing the Grid Control Machine Installation Once the system has rebooted, you can follow these steps to make sure LSF on the grid control machine is operating properly. 1. Log onto the grid control machine as one of the users you added to LSF using the

lspasswd command. These are usually lsfadmin, sasadm, sastrust, or sasdemo. 2. Open a DOS window by selecting Start->Programs->Accessories->Command Prompt 3. Run the command lsid. This should display the cluster name and the LSF master

machine name (i.e., grid control machine). 4. Run the command lshosts. This should display static information about the grid

control machine. 5. Run the command lsload. This should display dynamic information about the grid

control machine. 6. Run the command bsub sleep 100. This will submit a job to the cluster. 7. Run the command bjobs. This will display the job information. As you repeat this

command, you should see the job go from PEND, to RUN, to removed from the queue.

The following is sample output assuming the grid control machine (i.e., LSF master machine) is D15003.testgrid.com.

05Jan2007 of 41

Installing LSF on Grid Node Machines & Foundation SAS Clients 1. On the grid control machine, run lsf6.0_w2k_nt.exe. It is located on the 5th Platform

Suite for SAS CD.

05Jan2007 of 41

2. You should see the LSF InstallShield Welcome screen:


3. You will be asked where you want to unpack the LSF files:

This is usually a temp directory in the user’s home folder. If you do not want the temporary files there, enter a new directory. Press ‘Next’ to continue.

05Jan2007 of 41

4. The files should be unpacked and the progress should be displayed:

5. Next, the LSF Setup Welcome screen will appear.

Press ‘Next’ after reading the screen.

05Jan2007 of 41

6. Next, the Platform End User License Agreement is displayed.

Press ‘Agree’ or ‘Disagree’ after reading the agreement.

7. If you press ‘Agree’, the System Requirements screen is displayed.

Make sure the machine you are installing LSF upon meets the minimum requirements. If it does, press ‘Next’ to continue.

05Jan2007 of 41

8. Next, you are asked what action you want the install program to perform.

Since we are performing an install, make sure the ‘Install’ button is selected and press ‘Next’ to continue.

9. Next, you are asked whether you are creating a new cluster or adding to an existing

cluster.

Since we are installing on the grid node machine, we will be adding to an already existing cluster. Select ‘Add hosts to an existing LSF cluster’ and press ‘Next’ to continue.

05Jan2007 of 41

10. Next, since you adding to an existing cluster, LSF asks you for the name of one of the machines in that cluster so that it can retrieve configuration information.

Enter the fully qualified domain name of the grid control machine (LSF master machine) such as gridcontrol.domain.company.com. Press ‘Next’ to continue.

If LSF can locate the proper files, it will go directly to step 11. If LSF cannot access the registry on the grid control machine it may pop up the following error message:

Press ‘OK’ to continue to bring up the dialog to enter the network share.

05Jan2007 of 41

The network share dialog will then appear.

Enter the share path created when the grid control machine was installed (see step #15 of the Grid Control Machine install). This will allow LSF to read information out of the configuration files for the cluster.

11. Next, LSF needs to know the user accounts that will be used to administer the cluster.

If the domain LSF administrator(s) is not already filled in, enter the domain LSF administrator you created in the SAS pre-installation checklist (usually <domain>\lsfadmin where <domain> is your Windows domain name).

05Jan2007 of 41

12. Next, LSF needs to know the name of the computer you want to install LSF.

Enter the fully qualified domain name (FQDN) of one of your grid node machines (i.e., gridnode1.domain.company.com) and press the ‘ ->’ button. Using the FQDN will eliminate problems with multi-homed machines and clusters that include machines running operating systems other than Windows.

After pressing the ‘->’ button the name should appear in the right list box.

Repeat the previous step for all grid node machines you want to include in the cluster. When all grid nodes have been entered, press ‘Next’ to continue.

05Jan2007 of 41

13. Next, LSF wants to know where it should install its files.

Enter a new path or take the default and press ‘Next’ to continue.

14. Next, LSF needs a user account and password that it will use to run LSF services on

all machines.

Enter the LSF administrator ID you created in the pre-installation tasks. This is usually <domain>\lsfadmin. Press ‘Next’ to continue.

05Jan2007 of 41

15. Next, LSF will give you the option of receiving email whenever a job completes.

Leave the box unchecked and press ‘Next’ to continue.

16. LSF has completed getting all the information it needs and displays it in a summary

before starting the install.

Press ‘Next’ to start the install.

05Jan2007 of 41

17. The install may have a problem adding the new grid nodes in the lsf.cluster.cluster_name file on the grid control machine. If so, the following dialog is displayed:

If this occurs, you must log onto the grid control machine (LSF master machine) and copy the lsf.cluster.cluster_name.new file to the lsf.cluster.cluster_name file so that the other computers in the cluster will know about the new machines. Press ‘Continue’ to continue the install and you should see the install progress indicator.

05Jan2007 of 41

When the install is complete, LSF will display an installation results summary.

Each machine should be listed and the text should indicate the install and user setup succeeded. After checking that everything installed properly, press ‘Next’ to continue.

18. LSF has completed the install and displays its final screen.

Remove the check in the checkbox since you have already read the information when you set up the grid control machine. Press ‘Finish’ to continue.

05Jan2007 of 41

19. LSF will tell you that all new computers added to the cluster need to be rebooted and LSF on all machines in the cluster need to be restarted so that all computers in the cluster learn about the new computers.

20. If you installed Platform LSF on a machine that you want to be a grid node, you may

proceed to the next step. If however, you installed LSF on a machine you plan on just being a Foundation SAS grid client, you need to follow the instructions in Converting a Grid Node Machine to a Grid Client before you can continue. This is only for the case where you are going to run Foundation SAS for the purposes of submitting jobs to the grid without allowing that machine to participate as a grid node. Once you have made the necessary changes, proceed with the next step.

21. Reboot all new grid node machines and run the lsfrestart command on the grid

control machine so that all the existing machines in the cluster learn about the new machines. To run the command, start a DOS command prompt using Start->Programs->Accessories->Command Prompt and execute the command.

22. Finally, you must enable the cluster to work with the SAS Business Intelligence

Platform by following the instructions in the Supporting Grid Computing chapter in the SAS Intelligence Platform: Application Server Administrative Guide.

05Jan2007 of 41

Testing the Grid Node Installation Once the newly added machines have rebooted, you can follow these steps to make sure LSF is operating properly. 1. Log onto the grid control machine as one of the users that you added to LSF through

the lspasswd command. These are usually lsfadmin, sasadm, sastrust, or sasdemo. 2. Open a DOS window by selecting Start->Programs->Accessories->Command Prompt 3. Run the command lsid. This should display the cluster name and the LSF master

machine name (i.e., grid control machine). 4. Run the command lshosts. This should display static information about all the

machines in the cluster. 5. Run the command lsload. This should display dynamic information about all the

machines in the cluster. 6. Run the command bsub sleep 1000. This will submit a job to the cluster. Repeat this

command once for each node in the cluster. 7. Run the command bjobs. This will display the job information. As you repeat this

command, you should see the jobs being sent to the nodes in the cluster and going from PEND, to RUN, to removed from the queue.

The following is sample output of a heterogeneous cluster where the grid control machine (i.e., LSF master machine) is D15003.testgrid.com running the Windows operating system and the grid nodes are grid1.testgrid.com, grid2.testgrid.com, and grid3.testgrid.com, all running Linux. For a homogeneous cluster, all machines will be running the same operating system.

05Jan2007 of 41

Adding Grid Nodes or Foundation SAS Clients to the Cluster Adding new grid nodes or Foundation SAS clients to an existing cluster is identical to adding them to a new cluster. Follow the steps in the Installing LSF on Grid Node Machines & Foundation SAS Clients, specifying the new machines in step 12.

05Jan2007 of 41

Installing Grid Management Service (GMS) Grid Management Service is a daemon that is used by the Grid Manager Plug-in for the SAS Management Console to display grid information. After installing this service, you should be able to use the SAS Management console to view grid information. 1. Log onto the grid control machine with a user account that is an administrator on that

machine. 2. Run lsf6.0_gms_w2k.exe. It is located on the 1st Platform Suite for SAS CD.

05Jan2007 of 41

3. You should be presented with a window showing the extraction progress.

4. When extraction is complete, the Grid Management Service Installer screen is

displayed.


05Jan2007 of 41

5. Next the install asks you for the location to install the Grid Management Service files.


6. Next the installer asks for the port the Grid Management Service is going to use for

incoming connections from the SAS Management Console’s Grid Manager Plug-in.


05Jan2007 of 41

7. Next, the install will ask you for the user ID that will be used to run the Grid Management Service service programs.

You should enter the domain LSF administrator you created in the SAS pre-installation checklist.

8. The install has gathered all the information it needs and will present you with a summary before starting the install.


05Jan2007 of 41

9. The install will display the progress as it installs Grid Management Service.

10. GMS has completed the install and displays its final screen.

Press ‘Finish’ to complete the installation. NOTE: It is good idea to use the Windows’ ‘Services’ administration tool to check to make sure the ‘Grid Management Service GABD’ has started successfully.

05Jan2007 of 41

Post Installation Tasks Now that you have installed Platform Suite for SAS you still must enable the cluster to work with the SAS. To do this, please refer to the Supporting Grid Computing chapter in the SAS Intelligence Platform: Application Server Administrative Guide. This section is available online at http://support.sas.com/onlinedoc/913/docMainpage.jsp?topic=biasag.hlp/a003087200.htm

05Jan2007 of 41

Converting a Grid Node Machine to a Grid Client If you installed Platform LSF on a machine that you are going to run Foundation SAS for the purposes of submitting jobs to the grid without allowing that machine to participate as a grid node, you must set the number of job slots for that machine to zero so that no jobs will be assigned to the machine for processing. To change the job slots for a machine in the cluster, do the following: 1. Edit the lsb.hosts file in the LSF lsbatch configuration directory for your cluster on

the grid control machine. • On Windows this is usually

‘C:LSF_6.0\conf\lsbatch\<cluster_name>\configdir\lsb.hosts’. • On Unix this is ‘LSF_TOP/conf/lsbatch/<cluster_name>/configdir/lsb.hosts’.

2. In the ‘hosts’ section, add a new line for the grid client and set the maximum number

of jobs to 0.The easiest way to do this is to copy the line for the ‘default’ host, change the name, and change the MXJ parameter. This section looks like

Begin Host HOST_NAME MXJ r1m pg ls tmp DISPATCH_WINDOW #hostA () 3.5/4.5 15/ 12/15 0 () # Example #SPARCIPC () 4.0/5.0 18 16 () () # Example default ! () () () () () # Default End Host

For example, to add grid5.testgrid.com as a grid client, the resulting Host section would look like Begin Host HOST_NAME MXJ r1m pg ls tmp DISPATCH_WINDOW #hostA () 3.5/4.5 15/ 12/15 0 () # Example #SPARCIPC () 4.0/5.0 18 16 () () # Example default ! () () () () () # Default grid5.testgrid.com 0 () () () () () End Host

3. If you are not doing this as part of an install, you must restart LSF in order for it to pick up the changes. If you are doing this as part of an install, you will restart LSF at the end of the install, so the restart does not need to be done here.

05Jan2007 of 41

LSF Quick Reference Command Description

lsid Displays version number, cluster name, and the master host name. Useful to see if the grid daemons are running and if running in SAS mode.

lshosts Displays information about the hosts recognized by LSF along with their static resource information.

lsload Displays the dynamic resource information for the hosts in the grid. lsadmin Administrative tool for LSF available to LSF administrators. Useful

subcommands are reconfig Restarts all LIMs in the cluster to read any changes

in the configuration files. limstartup Starts LIM on the local host limrestart Restarts LIM on the local host resstartup Starts RES on local host resrestart Restarts RES on local host

lsfstartup Starts the LIM, RES, sbatchd, and mbatchd daemons on all hosts in the cluster. Must be run as root and all hosts must be running rsh or ssh daemons.

lsfrestart Restarts the LIM, RES, sbatchd, and mbatchd daemons on all hosts in the cluster. Must be run as root and all hosts must be running rsh or ssh daemons.

lsfshutdown Shuts down the LIM, RES, sbatchd, and mbatchd daemons daemons on all hosts in the cluster. Must be run as root and all hosts must be running rsh or ssh daemons.

bhosts Displays batch information about all hosts in the cluster. bjobs Displays information about current user’s LSF jobs bhist Displays historical information about jobs. Useful parameters are

-p | -r | -d | -a Displays information about specific jobs (pending, running, done, or all).

-l Display in long format. -u <user> | all Displays job for specified or all users. <job ID> Displays only specified job information.

badmin Administrative tool for LSF’s batch processing facility available to LSF administrators. Useful subcommands are reconfig Reconfigures the batch facility without

restarting sbatchd or mbatchd to read any changes in the configuration files.

hstartup Starts sbatchd on the local host hrestart Restarts sbatchd on the local host mbdrestart Restarts mbatchd hclose <host> Closes a host preventing it from running jobs. hopen <host> Opens a host to allow it to run jobs.

bsub Submit a job to the grid. Useful parameters are -I Interactive job. Output displayed on local

terminal. -m Submit to a specific host. -R “res_req” Submit with specified resource

step-by-step installation of platform suite for sas on...

Documents