Deploying BMC AMI Platform

This topic describes how to download and deploy BMC AMI Platform.

Before you begin

Make sure you've reviewed the System requirements topic before you begin this process.

Downloading BMC AMI Platform from EPD

1. Log in toElectronic Product Distribution - BMC Mainframe.
   Warning

   Important

   Access to the EPD site depends on your company’s license entitlements.

2. Select All/Trial Products.

3. Search for and select BMC AMI Platform, and then download the BMC-AMI-PLATFORM-2.0.00.zip file.

To deploy BMC AMI Platform on Kubernetes

1. Log in to the primary manager node of the Kubernetes cluster with a user who has sufficient permissions to run kubectl commands and copy the downloaded zip file to your preferred folder, such as /home.
2. Extract all deployment files into this directory by using the following command:
   unzip BMC-AMI-PLATFORM-2.0.00.zip -d BMC-AMI-PLATFORM-2.0.00
   The path should be /<extracted_dir>/BMC-AMI-PLATFORM-2.0.00.

3. Provide permission to the folder by using the following command:
   chmod -R 777 /<extracted_dir>/BMC-AMI-PLATFORM-2.0.00
4. Verify that the following files and directories are present and correctly structured:
   Root structure

       ansible-playbook/
           site.yml
   
       config/
           nfs_volume_paths.yaml
   
       helm_charts/
           01-helm-service-registry/
           02-helm-data-service/
           03-helm-milvusdb-service/
           04-helm-security-service/
           05-helm-zosconnector-service/
           06-helm-core-common/
           07-helm-amiai-chart/
           08-helm-api-gateway/
           09-helm-swarm-redis-cluster/
           10-datastore-init-scripts/
           11-helm-elasticsearch/
           12-helm-core-notification/
   
       static/
           namespaces.yaml
           secrets/
       setup-script.sh  
       scripts/
          code-conversion.sh
          code-explain.sh
          granite.sh
          llama.sh
          mixtral.sh
          oi-rc.sh
          undeploy_granite.sh
          undeploy_llama.sh
          undeploy_mixtral.sh

To run the setup and deployment script

1. Run the following commands:

   Warning

   Important

   You must have a user who has sufficient permissions to run kubectl commands.

   cd /<extracted_dir>/BMC-AMI-PLATFORM-2.0.00/
   
   ./setup-script.sh
    
2. Use the following fields to assist you during the configuration process:

   Field	Description
   Kubeconfig file	Select how to configure the Kubernetes access: Use the current Kubernetes context (default: ~/.kube/config) Provide the path to a kubeconfig file. Paste kubeconfig YAML (end input with a single line containing: EOF) To configure the kubeconfig file, you have the following three options: (Recommended) To use your current kubectl context, enter 1. If you’re using a different kubeconfig file, enter 2 and provide its full path. To paste the kubeconfig content manually, enter 3.
   Registry and project	Based on where you plan to download the BMC AMI Platform images, select one of the following values: Y—Get default values to use the BMC container (containers.bmc.com) and BMC project name. N—Add your own values to the following fields: Docker registry URL Project name
   Docker image registry user name	Enter the Docker image registry user name. It is visible in the UI.
   Docker image registry password	Enter the Docker image registry password. It is visible in the UI.
   Docker Hub registry for the Uptrace image	To use the Uptrace image in BMC AMI Platform, you must provide the following Docker Hub credentials to pull it from the docker.io registry: Provide Docker Hub registry credentials (required for pulling the uptrace/uptrace:2.0.1 image). Configure Docker Hub credentials for Uptrace? [Y/N]: Docker Hub Registry URL [docker.io]: Docker Hub Username: Docker Hub Password:
   TLS certificate path	Enter the TLS certificate path.   For more information, see SSL and TLS requirements.
   TLS key path	Enter the TLS key path.   For more information, see SSL and TLS requirements.
   Password for default users	Enter the password for all the default users in the system. A summary table of the relevant fields is displayed for your confirmation.
   NFS server (IP or Hostname)	Enter your NFS Server host name or IP.
   NFS server path	Enter your NFS server path details.
   Local mounted path	Enter local mounted path (such as /mnt/data).
   BMC AMI Platform machine SSH key	Enter your SSH key (such as /root/.ssh/id_rsa).
   BMC AMI Platform application host machine (IP or host name)	Enter the IP or host name of the machine where you run the deployment script.
   BMC AMI Platform machine user	Enter the machine user name (such as root or ubuntu). A summary table of the relevant fields is displayed for your confirmation.
   GPU nodes names	Enter the names of your GPU nodes as displayed in the available nodes list. If you have more than one GPU node, use a comma between the node names. You can press Enter to skip GPU node labeling if not required. A confirmation message [Y/N] appears.
   NVIDIA GPU operator	To deploy the NVIDIA GPU operator with the current script, enter Y. To deploy it on your own, enter N. A summary table of the relevant fields is displayed for your confirmation.
   Cloud or On-prem deployment	For cloud deployment, enter 1. For On-prem deployment, enter 2. If you enter 2, three requests appear as follows: Enter the address (DNS name, host name, or IP) for Cloud UI access. Information Example example.bmc-ami-platform.com or alb-12345.elb.amazonaws.com Use a custom port instead of the default 443? [Y/N] Proceed with this Platform UI URL? [Y/N] After you select your deployment type, the following question appears: Run Ansible playbook to deploy Helm releases? To complete the process, enter Y. To cancel the entire process, enter N.

After your script has finished running, you get the following message confirming that the task was completed successfully:

To verify the deployment

1. Verify that the pods are running successfully under the name spaces by using the commands in the following examples:

Before you proceed, make sure you have the following:

1. Download Jobs—Confirm that all three download jobs are completed.
2. Load Embedding Job—Verify that the load embedding job has finished successfully.