LSF Installation Quick Reference

Platform Computing Corporation


Overview

Follow these instructions to extract the software from the distribution medium, then use the lsfsetup command to install and configure LSF. lsfsetup prompts for all information needed to install LSF. Most prompts include a default value, displayed in square brackets. If you press [RETURN], that value is used. If you type a question mark ([?]) at any prompt, a short help message is displayed.

Privileges Needed for Installation

These instructions assume you have some UNIX system administration experience, your environment has uniform user names (UID), a shared file system structure across your network, and the network is behind a firewall. You must have superuser privileges to install LSF. Your Network File System (NFS) mounted directories must be exported and mounted setuid (client root access).

Reading the Distribution Medium

LSF is available on a variety of distribution media (ftp net access, CD-ROM, or tape).

The LSF files are compressed tar archives for each host type supported by LSF. The archives are named by LSF version and host type, for example lsf3.0_sparc-sol2.tar.Z. Extract the desired archives into a temporary directory, for example /var/tmp. Uncompressing the archive creates a distribution directory, for example lsf3.0_sparc-sol2. The distribution directory takes between 50MB to 120MB for each architecture. You need to plan your space requirements accordingly.

Selecting the Desired LSF Components

LSF Suite V3.0 includes the following components: LSF Base, LSF Batch, LSF JobScheduler, and LSF MultiCluster. The distribution file for each host type contains all the components. During installation, specify which components you want to run.

The default installation is LSF Batch together with LSF Base. LSF Base is a prerequisite for all other LSF components.

Installing LSF with lsfsetup

The following examples install the LSF binaries for Solaris 2.x on the file server. The distribution directory is called lsf3.0_sparc-sol2 and requires approximately 40.2MB of disk space:

Installing the First Host Type

1. Log in to the file server host as root. Change to the temporary directory. Extract the Solaris 2.x version of the LSF software from the distribution tape. Uncompress the distribution file. Change directory to the distribution directory.

% cd /var/tmp
% zcat lsf3.0_sparc-sol2.tar.Z | tar xf -
% cd lsf3.0_sparc-sol2

2. Run the lsfsetup command:

% ./lsfsetup

3. Choose '1. Install for the First Time'. Then select one or more of the following components to install: '1. Batch (default)', '2. JobScheduler', '3. MultiCluster', or '4. Base-only' by typing in the corresponding number. If you select more than one component, make sure you separate them with commas:

Please enter your choice(s) separated by comma (e.g., 1,3)? [1] 1

4. Select the installation directories. For the host independent installation directory, enter /usr/local/lsf. For the host type dependent installation directory, enter /usr/local/lsf/mnt/SOLARIS. Enter the name of the LSF cluster, and the login name of the LSF administrator:

LSF_INDEP? [? for help] [/usr/local/lsf] /lsf/local/lsf
...
LSF_MACHDEP? [? for help] [/lsf/local/lsf] /lsf/local/lsf/mnt/SOLARIS
...
LSF_CLUSTER_NAME? [? for help] test_cluster
...
LSF_MANAGER? [? for help] [lsf] lsf

5. Review and modify the settings as needed. Choose '1. List Current Settings' to review the configuration settings. If you wish to change any setting or see the help text for any setting, choose '2. Change Current Settings'.

6. Choose '3. Install the Software Now'. This copies all the LSF files into their final locations and creates a default set of cluster configuration files. The lsfsetup command is installed into the LSF_SERVERDIR directory. You can use lsfsetup later to maintain LSF.

Installing Additional Host Types

For the following example, the HP-UX version of LSF is installed on the file server. The distribution directory is called lsf3.0_hpux and requires approximately 31.3MB of disk space.

1. Log in to the file server host as root. Change to the directory where you want the distribution directory created. Extract the HP-UX version of the LSF software from the distribution medium, in this example a tape. Uncompress the distribution file. Change directory to the HP-UX distribution directory.

% cd /var/tmp
% zcat lsf3.0_hpux.tar.Z | tar xf -
% cd ./lsf3.0_hpux

2. Copy the LSF configuration file from the Solaris 2.x distribution directories with the following command:

# cp /usr/local/lsf/mnt/SOLARIS/etc/lsf.conf .

3. Run the lsfsetup command:

% ./lsfsetup

4. Choose '2. Install Another Host Type (for a mixed cluster)'.

5. The lsfsetup command asks whether you want to base your configuration on an existing lsf.conf file. Answer yes, and enter the name of the lsf.conf file you copied from the Solaris 2.x directories.

6. Enter /usr/local/lsf/mnt/HPUX as the directory for machine-dependent files. The other installation settings are automatically updated.

7. Choose '3. Install the Software Now'. The lsfsetup command installs all the components necessary for the additional host type.

Additional Steps on Each LSF Server Host

Each server host in the LSF cluster must be setup as follows:

1. Create a /usr/local/lsf/mnt directory and NFS mount the /usr/local/lsf/mnt directory from the file server. You must not mount this directory with the nosuid flag; some LSF programs require setuid permission.

2. Create the symbolic links appropriate for the host type. For example, on a Solaris 2.x host:

# ln -s /usr/local/lsf/mnt/SOLARIS/{bin,etc,lib} /usr/local/lsf

3. Log in as root and run the lsfsetup command.

% /usr/local/lsf/etc/lsfsetup

4. Choose '6. Host Setup'.

lsfsetup creates a symbolic link to the lsf.conf file (used by LSF daemons and applications to find the LSF configuration files and working directories) in LSF_ENVDIR (/etc by default). It adds the LSF services to the /etc/services file if they are not defined in NIS or NIS+ and configures the system so that the LSF daemons are started when the system boots.

Initial Configuration

Once the file server host is completely set up, you need to configure the system parameters. These parameters tell the LSF daemons about the definition of your cluster by defining hosts, host types, and resources in your cluster. You can use default settings for most parameters.

1. Log in as root and run the lsfsetup command.

% /usr/local/lsf/etc/lsfsetup

2. Choose '5. Configure LSF Cluster'. lsfsetup will prompt you to enter or confirm the location of your lsf.conf file. Then it will enter the configuration sub-menu.

3. Choose '4. View/Add/Delete Currently Defined Host Types'. Use the menu to include the host types you have in your cluster. When you are finished, select 'q. Done Modifying Host Types'.

4. Choose '2. View/Add/Delete/Modify Currently Defined Host Models'. Use the menu to add host model specifications and CPU factors for the systems in your cluster. When you are finished, select 'q. Done Modifying Host Models'.

5. Choose '1. View/Add/Delete/Modify Currently Configured Hosts'. From the sub-menu, choose '2. Add Hosts To LSF Configuration'. The lsfsetup command prompts for a host name and then runs the vi editor. Edit the appropriate line for the host and set the host type, model, load thresholds, and resources as required.

You can add more than one host at this time. Enter the host information for all hosts in the cluster. The master LIM and mbatchd daemons run on the first available host in this list, so you should put reliable server hosts first.

When you are done adding hosts select 'q. Done Modifying Host Configuration'.

6. Choose '5. Configuration File Error Checking'. This option runs a check of the LSF configuration files, and reports any problems found.

Select 'q. Quit' to exit the configuration sub-menu and again to exit lsfsetup.

Note
lsfsetup creates a default LSF Batch configuration, including a set of batch queues. You do not need to change any LSF Batch files to use the default configuration. After you have the system running, you may want to reconfigure LSF Batch.

Installing the FLEXlm License Key

LSF uses the FLEXlm license management software from Globetrotter Software. FLEXlm supports network-wide permanent licenses as well as limited time DEMO licenses. You must have a license key to run LSF. These instructions only show using a DEMO license.

Getting a DEMO License

Contact your LSF vendor. Tell them components of LSF you want to run. They will give you a time limited license key(s) with the desired feature(s) enabled.

Creating the license.dat File

Store the license key(s) in a a file named license.dat in the distribution directory. The LIM on every host need to read the license key, so this file should be available on all LSF hosts. It can be shared using NFS or you can install a copy on each host.

Installing a DEMO license

If you are installing LSF with a DEMO license, you do not need to run the license server daemons. Each LSF host reads the license file directly. Until the license expires, LSF will run without trying to contact a server.

1. Change directory to the distribution directory. Run the lsfsetup command.

% cd /var/tmp/lsf3.0_sparc-sol2
% ./lsfsetup

2. Choose '4. License Management'. From the next menu, choose '2. Install a New License From the LSF Vendor'.

3. From the next menu choose '1. Set up a DEMO license'.

It looks for the license key in a license.dat file in the current and parent directories. If the file is not found, it prompts you to enter the full path name of the file containing your DEMO license key.

4. You are then prompted for a location in which to store the installed license file. Choose '2. Use $LSF_CONFDIR/license.dat' and type 'y' to commit the changes.

lsfsetup checks the license.dat file to confirm the licensed components. It then prompts for permission to enter appropriate features in the 'FEATURES=...' line in the lsf.cluster.cluster file.

lsfsetup installs the key in the FLEXlm license file, creating it if necessary. It updates the lsf.conf file to include the full path name of the license file.

Checking the LSF Configuration

Before you can start any LSF daemons, you should make sure that your cluster configuration is correct. To confirm the LSF configuration, you can run the ckconfig option of the lsadmin and badmin commands.

Log into the first host listed in lsf.cluster.cluster as the LSF administrator and check LIM configuration with the following command:

% lsadmin ckconfig -v

To check the LSF Batch configuration files, LIM must be running on the master server. Log in as root and start LSF_SERVERDIR/lim. Wait a minute and then run the lsid program to make sure LIM is available. Then run the following command to check the LSF Batch configuration:

% badmin ckconfig -v

Starting the LSF Daemons

The LSF daemons must be run by root on every host in the cluster. When LSF is first configured, the daemons are started by hand.

1. Start the LSF daemons by running LSF_SERVERDIR/lsf_daemons start and use the ps command to make sure that res, lim and sbatchd have started.

Under normal operation, the daemons are restarted automatically each time the host boots. The steps required to set up daemons are different under different versions of UNIX. In all cases, the LSF daemons should be started after NFS and all other networking daemons.

Testing the LSF Cluster

Testing consists of running a number of LSF commands and making sure that correct results are reported for all hosts in the cluster. Log in as a non-root user. Your PATH must include the LSF user binaries (LSF_BINDIR as defined in /etc/lsf.conf). The following steps may be performed from any host in the cluster.

1. Check cluster name and master host name:

% lsid

2. The lsinfo command displays cluster-wide configuration information.

% lsinfo

3. The lshosts command displays configuration information about your hosts:

% lshosts

4. Check the current load levels:

% lsload

5. The lsgrun command runs a UNIX command on a group of hosts:

% lsgrun -v -m "hostA hostD hostB" hostname

6. The bhosts command lists the LSF Batch server hosts in the cluster:

% bhosts

7. Submit a job to the default queue:

% bsub sleep 60

8. Check available queues and their configuration parameters:

% bqueues

9. Check job status:

% bjobs

Documentation

Documentation for LSF consists of this reference and the LSF Administrator's Quick Reference, the LSF User's Quick Reference, the LSF Administrator's Guide, the LSF User's Guide, the LSF JobScheduler User's Guide, the LSF Programmer's Guide, the LSF man pages, and the xlsbatch, xlsmon, xbsub, and xbcal on-line help.

Note
To install LSF on Windows NT, please refer to the instructions in the file README.NT in the distribution directory. The Windows NT version is supplied as a ZIP archive. You need WinZip, PKZIP, or similar applications to extract the files and directories.


doc@platform.com

Copyright © 1994-1997 Platform Computing Corporation.
All rights reserved.