Creating an LSF Workgroups cluster after installation

Create an IBM® Spectrum LSF® Suite for Workgroups 10.1 cluster in IBM Spectrum Cluster Foundation Community Edition using the LSF Workgroups (LSF_Workgroups) cluster template.

1. Log in to the IBM Spectrum Cluster Foundation Community Edition Web Portal as a system administrator.
2. Create a cluster from the published LSF Workgroups (LSF_Workgroups) cluster template:
   1. Go to the Clusters tab, and select Clusters.
   2. Click Create.
   3. Select the LSF_Workgroups cluster template and click Create.
   4. Click the Servers tab, set the Minimum Number of Servers for the LSFCompute tier.
      Restriction: The minimum number of servers in the LSFMaster tier must be set to 0 and cannot be changed.
   5. Click the User-defined variables tab, set up your cluster configurations.

      ENABLE_LSF_ADDONS_HA

      Important: To enable high availability in LSF, set the ENABLE_LSF_ADDONS_HA variable to Y only if you enabled high availability during IBM Spectrum Cluster Foundation Community Edition installation. If you did not enable high availability during IBM Spectrum Cluster Foundation Community Edition installation, set the ENABLE_LSF_ADDONS_HA variable to N.

      LSF_ADMIN_USER

      If high availability is enabled, make sure that the user that was created on the standby management node is the value specified for LSF_ADMIN_USER.

   6. Click Create to provision your LSF cluster. Provisioning an LSF cluster can take several minutes.
3. If you installed the product with enabling high availability, run the following command on the management node to complete your high availability configurations.

   # /opt/pcm/bin/setup_pac_ha

4. From the command line, source your environments.
   • Source IBM Spectrum Cluster Foundation Community Edition:

     # source /opt/pcm/bin/pcmenv.sh

   • Source IBM Spectrum LSF:

     # source /shared/ibm/lsf/conf/profile.lsf

   • Source IBM Spectrum LSF Application Center:

     # source /opt/pac/profile.platform

5. Verify that your cluster was provisioned successfully.
   1. In the Clusters page, verify that your cluster is in Active state. Note that while provisioning, the cluster state is set to Provisioning before changing to Active.
      If any errors occur during cluster provisioning, refer to the appropriate log file to troubleshoot. The following LSF cluster log files are found on the management node in the /tmp directory, they include:

      Table 1. LSF cluster log files on the management node

      Log File (/tmp)	Script Template
      postMasterLSF.log	postMasterLSF
      postPAC.log	postMasterPAC
      postPMPI.log	postMasterPMPI
      PostLS.log	postMasterLS

      The following LSF cluster log files are found on servers:

      Table 2. LSF cluster log files on the servers

      Log File (/tmp)	Script Template
      postComputeLSF.log	postComputeLSF
      postPMPI.log	postComputePMPI

   1. Verify that the PAC Web Portal is accessible.
      1. While the cluster is selected on the Clusters page, in the bottom pane, click the Summary tab.
      2. In the Summary tab, locate the PAC URL field and open the URL provided in a web browser.
      3. Log in to the PAC Web Portal as an LSF administrator. By default, the user name and password are both lsfadmin.
      4. From the PAC Web Portal you can see the status of your LSF cluster.
   2. To verify your LSF cluster status from the command line, do the following steps:
      1. Navigate to the /shared/ibm/lsf directory.

         # cd /shared/ibm/lsf

      2. Run various LSF status commands to see information about the LSFCluster cluster.

         # . conf/profile.lsf 
         # lsid 
         # lshosts 
         # bhosts

         If the lsid, lshosts, and bhosts commands provide information about the LSF cluster, the LSF cluster is running correctly.

         If high availability is enabled, information about both management nodes is available. For example, information about the primary and secondary management node is displayed when the bhosts command is run.

         # bhosts
         HOST_NAME  STATUS  JL/U  MAX  NJOBS  RUN  SSUSP  USUSP   RSV
         <smn>      ok      -     1    0      0    0      0       0
         <mn>       ok      -     1    0      0    0      0       0