Overview

IBM Platform LSF Server v9.1.3

IBM Platform LSF is a powerful workload management platform for demanding, distributed High Performance Computing (HPC) environments. It provides a comprehensive set of intelligent, policy-driven scheduling features that enable you to utilize all of your compute infrastructure resources and ensure optimal application performance.

A LSF Server host is a server that is capable of submitting and executing jobs. They cannot be selected as LSF Master incase the designated LSF Master goes down.

More information available at the IBM Knowledge Center

Please note that the LSF charms (LSF Storage, LSF Master and LSF Server) will create LSF cluster having same host type (either x86 or ppcle). LSF supports different host types to be part of LSF Cluster. But the LSF charms as of now supports same host type model only (So all the hosts ie master, server should have same machine architecture either x86 or ppcle).

Usage

This charm will not download or require any LSF product binaries as that is being handled by the IBM Platform LSF Storage charm. The LSF installation/configuration files will be shared when a relation is added between LSF Storage and LSF Server (using NFS).

Deploy

This charm will deploy only LSF Server. Functionally this charm will do nothing. To have a working LSF Cluster, you need to add relations between LSF Server and LSF Storage and between LSF Server and LSF Master.

To deploy IBM Platform LSF Server charm, follow the below steps:

1) Deploy the LSF Storage charm

juju deploy ibm-platform-lsf-storage \
  --resource ibm_lsf_installer=</path/to/installer.tar.Z> \
  --resource ibm_lsf_installation_x86=</path/to/distribution.tar.Z> \
  --resource ibm_lsf_entitlement_file=</path/to/entitlement.dat>

Note: The resource for getting the Installation/Distribution is different for x86 and Power, so if you are deploying on x86 machines, use resource name as ibm_lsf_installation_x86 and for Power as ibm_lsf_installation_ppcle

2) Deploy the LSF Master charm

juju deploy ibm-platform-lsf-master

3) Deploy the LSF Server charm

juju deploy ibm-platform-lsf-server

4) Add a relation between LSF Storage and LSF Master and LSF Storage and LSF Server to share the LSF installation and configuration files

juju add-relation ibm-platform-lsf-storage ibm-platform-lsf-server
juju add-relation ibm-platform-lsf-storage ibm-platform-lsf-master

5) Add a relation between LSF Master and LSF Server.

juju add-relation ibm-platform-lsf-master ibm-platform-lsf-server

Add units of LSF Server

juju add-unit ibm-platform-lsf-server

This will add one more LSF Server host to your existing LSF Cluster.

Remove units of LSF Server

juju remove-unit <unit name of the LSF Server you want to remove>

This will remove the unit from the existing LSF Cluster. (LSF conf files will be updated and LSF files will be unmounted)

Removing Relations

A LSF Server is related to IBM Platform LSF Storage charm as well as IBM Platform LSF Master charm. In case you want to remove the relation between these, refer to the below steps:

juju remove-relation ibm-platform-lsf-master ibm-platform-lsf-server

This will remove the server host information from the lsf configuration files and the server host will no longer be part of the existing LSF Cluster. The LSF Server Daemons will be stopped.

juju remove-relation ibm-platform-lsf-storage ibm-platform-lsf-server

This will unmount the LSF shared installation files.

Verify the LSF Cluster

Once the IBM Platform LSF Server is deployed successfully and relation is established between LSF Storage and LSF Server and LSF Master and LSF Server, you can verify your LSF cluster by running some lsf commands as mentioned below:

• Login into the machine where LSF Server is installed as lsf administrator user lsfadmin

• Export the lsf profile path: . /usr/share/lsf/conf/profile.lsf

• Go to the path /usr/share/lsf/9.1/linux2.6-glibc2.3-x86_64/bin. The distribution folder will change based upon Operating System, for Power the path would be /usr/share/lsf/9.1/linux3.10-glibc2.17-ppc64le/bin and you can issue commands like lsid, lshosts, bhosts to verify your cluster is up and fine.

• For submitting jobs you can issue command bsub <job>, for example: bsub pwd

IBM Platform LSF Information

(1) General Information

Information on IBM Platform LSF available at the IBM Knowledge Center

(2) Contact Information

For issues with this charm, please contact IBM Juju Support Team jujusupp@us.ibm.com

(3) Known Limitations

This charm is dependent on LSF Storage charm which makes use of Juju features that are only available in version 2.0 or greater.