Installing variable use licensing on Linux

Track prepaid usage of cloud resources in your cluster by installing and configuring IBM Spectrum LSF Suite for HPC on Linux® hosts to upload usage metrics to IBM Cloud Private. With variable use licensing for the cloud, CPU core and socket, GPU socket, and host count metrics are uploaded through an external load index monitor (ELIM) in IBM Spectrum LSF Suite for HPC to a metering service in IBM Cloud Private. You can then audit IBM Spectrum LSF Suite for HPC resource usage from the metering dashboard in IBM Cloud Private.

1. Log on to the LSF management host as root.
2. Set your environment:
   • For csh or tcsh: % source LSF_TOP/conf/cshrc.lsf
   • For sh, ksh, or bash: $ . LSF_TOP/conf/profile.lsf
3. Install the variable use licensing package that you downloaded from Passport Advantage. This package specifies licensing terms and extracts the files to install prepaid monitoring in your cluster.
   1. Go to the patch install directory: cd $LSF_ENVDIR/../10.1/install/
   2. Copy the patch file to the installation directory: $LSF_ENVDIR/../10.1/install/
   3. Extract the patch file.
   4. Run zcat <patch tar file>|tar xvf -.
   5. Copy elim.icp.uploader to the /opt/ibm/lsfsuite/lsf/10.1/<system version>/etc directory.
4. Generate an API key for the IBM Cloud Private 3.1 metering service.
   1. Log on to the IBM Cloud Private boot node and obtain the cluster configuration details according to step 3 in https://www.ibm.com/support/knowledgecenter/SSBS6K_3.1.0/manage_cluster/cfc_cli.html.
   2. Using the following cloudctl commands, log in as the administrator (admin) and generate an API key, substituting the current node's IP address for <my_icp_IP>:
      1. cloudctl login -a https://<my_icp_IP>:8443 -u admin -p admin -n default --skip-ssl-validation
      2. cloudctl iam service-id-create spectrumserviceid-default-1 -d "SPECTRUM serviceid in default namespace, number 1"
      3. cloudctl iam service-api-key-create spectrumapikey-default-1 spectrumserviceid-default-1

      The output of the last command displays the API key, which is similar to the following: OiiJ1q3gjIDk2JJGxHK6xeZwU5mrIHyu2nUv38SzrQSA

      Note: Save the API key to add to the icpup.json file in the following step.
5. Configure IBM Spectrum LSF to upload usage data to IBM Cloud Private.
   1. Log in to the LSF management host as root, edit the $LSF_ENVDIR/lsf.shared configuration file, and add the following line to the Resource section:

      totalcores Numeric 15 Y (Total CPU cores in the cluster)

   2. Edit the $LSF_ENVDIR/lsf.cluster.* configuration file and add the following line to the Resource Map section

      totalcores [all]

   3. Create the new $LSF_ENVDIR/icpup.json file and insert the following contents, replacing <my_icp_IP> and <my_API_key> with actual values:

       {
          "icpUrl": "https://<my_icp_IP>:8443",
          "apiKey": "<my_API_key>",
          "logLevel": "LOG_INFO"
         }

      To set the log level, change the logLevel parameter to any of the following values: LOG_INFO, LOG_WARNING, LOG_ERR, or LOG_DEBUG.

   4. Restart lim on LSF management host.

      For IBM Spectrum LSF Suite for HPC clusters running version 10.2, run the systemctl restart lsfd command.

   5. Check the ELIM status.

      For IBM Spectrum LSF Suite for HPC clusters look in the /opt/ibm/lsflogs/elim.icpup.log.<hostname> file.

   6. Log in to the IBM Cloud Private console and navigate to the Metering page to see the uploaded usage data. For more information, see https://www.ibm.com/support/knowledgecenter/SSBS6K_3.1.0/manage_metrics/metering_service.html.
      The following metrics are uploaded to IBM Cloud Private:

      • Total Hosts: Total number of available and closed hosts in the cluster. The other metrics are for these reported hosts and do not include unavailable hosts.
      • Total Cores: Total number of CPU cores on the reported hosts.
      • Total GPUs: Total number of GPU processors on the reported hosts.
      • Total Sockets: Total number of CPU sockets on the reported hosts.

      CPU core and GPU socket data for the previous two days are immediately uploaded to the Metering page. Host and socket usage is uploaded from the current time. The system collects metrics every 5 minutes and uploads to the metering service every 15 minutes.