Skip to content

Deploy Additional Packages

This section explains how to download and deploy additional software packages and container images on the cluster nodes using Omnia local repositories.

Overview

Omnia supports deploying additional packages and container images at two stages:

  1. During first-time deployment -- Packages are included in the cluster node images before provisioning.
  2. After cluster provisioning -- Packages are added to running nodes post-deployment.

For adding extra RPM repositories (not individual packages), see Deploy Additional Repositories.

Prerequisites

Configure additional container images from a user registry to the local repository. Omnia supports configuring additional container images from specified user registries to the Omnia Local Repository, so that these images are available and can be pulled by Service Kubernetes Cluster nodes as per requirement. User registries may be hosted either on OIM or on an external server, and both HTTP and HTTPS registries are supported.

After the registry is ready, mention the inputs in local_repo_config.yml. See Local Repo Config.

Steps

Step 1: Add additional_packages to software_config.json

Open /opt/omnia/input/project_default/software_config.json and add the additional_packages entry under the softwares list. Also define the additional_packages section to specify which node roles should receive the additional packages.Save the software_config.json file.

Sample software_config.json with additional_packages enabled:

{
  "cluster_os_type": "rhel",
  "cluster_os_version": "10.0",
  "repo_config": "partial",
  "softwares": [
    {"name": "default_packages", "arch": ["x86_64","aarch64"]},
    {"name": "additional_packages", "arch": ["x86_64","aarch64"]}
  ],
  "additional_packages": [
    {"name": "slurm_control_node"},
    {"name": "slurm_node"},
    {"name": "login_node"},
    {"name": "login_compiler_node"},
    {"name": "service_kube_control_plane"},
    {"name": "service_kube_node"},
    {"name": "os"}
  ]
}

Note

  • Ensure the arch list matches your cluster architecture(s).
  • To install debug packages, also add: {"name": "admin_debug_packages", "arch": ["x86_64", "aarch64"]}

Step 2: Configure additional_packages.json

Update the additional_packages.json file at /opt/omnia/input/project_default/config/<architecture>/rhel/10.0/ with the required packages and images.Save the additional_packages.json file.

Each entry needs:

  • For RPM packages: package, type: "rpm", and repo_name
  • For container images: package, type: "image", and tag or digest

Sample additional_packages.json:

{
  "additional_packages": {
    "cluster": [
      {
        "package": "fuse-overlayfs",
        "type": "rpm",
        "repo_name": "appstream"
      },
      {
        "package": "sssd",
        "type": "rpm",
        "repo_name": "baseos"
      },
      {
        "package": "quay.io/strimzi/kafka-bridge",
        "type": "image",
        "tag": "0.33.1"
      },
      {
        "package": "172.16.0.254:7000/ubuntu/squid",
        "type": "image",
        "tag": "latest"
      }
    ]
  },
  "os": {
    "cluster": [
      {
        "package": "podman",
        "type": "rpm",
        "repo_name": "appstream"
      },
      {
        "package": "curl",
        "type": "rpm",
        "repo_name": "baseos"
      }
    ]
  }
}

Note

All container images specified in additional_packages.json are configured in the Omnia local repository and can be pulled on all Service Kubernetes Cluster nodes.

Architecture-Specific Guidelines

The additional_packages feature has architecture-specific functional group support:

  • x86_64: slurm_control_node, slurm_node, login_node, login_compiler_node, service_kube_control_plane, service_kube_control_plane_first, service_kube_node, os
  • aarch64: slurm_node, login_node, login_compiler_node, os

Architecture-specific JSON files are located at:

  • /opt/omnia/input/<project_name>/config/x86_64/rhel/<version>/additional_packages.json
  • /opt/omnia/input/<project_name>/config/aarch64/rhel/<version>/additional_packages.json

Warning

  • Kubernetes-related groups (service_kube_*) and slurm_control_node are only supported on x86_64.
  • Include only supported functional groups for each architecture.
  • Review validation logs at /opt/omnia/log/core/playbooks/validation_omnia_<project_name>.log for warnings.

Deploy Additional Packages During First-Time Deployment

  1. Configure software_config.json and additional_packages.json as described above.

  2. After the local repositories are created, build the cluster node images and PXE boot the nodes using the images:

  3. Build images: Build Cluster Images

  4. Discover nodes and PXE boot: Discover Nodes

Deploy Additional Packages After Cluster Provisioning

  1. Configure software_config.json and additional_packages.json as described above.

  2. Re-run the local_repo.yml playbook to download the new packages/images to the Pulp container.

  3. After the local repositories are updated:

  4. Install RPM packages on the required nodes:

    Run on: compute node
    dnf install <package-name>

  5. Pull container images on the required nodes (see Pulling images from a user registry via Pulp below).

  6. Verify the installed packages/images:

    Run on: compute node
    dnf list installed <package-name>
crictl images

Pulling Images from a User Registry via Pulp on a Service Kubernetes Cluster

When the container images from a user registry are specified in the additional_packages.json, running local_repo.yml thereafter uploads those images to the configured Pulp registry.

After this synchronization:

  • All cluster nodes must pull images from Pulp, not directly from the user registry.
  • This enables centralized image management and supports offline or air-gapped environments.

Example: Image defined in additional_packages.json:

"additional_packages": {
  "cluster": [
    {
      "package": "100.10.0.76:3445/library/nginx",
      "type": "image",
      "tag": "1.25.2-alpine-slim"
    }
  ]
}

In this example, 100.10.0.76:3445 is the user registry. Omnia syncs the image to the Pulp registry. Cluster nodes must subsequently pull the image from Pulp.

Retrieve the Pulp Registry Endpoint

On the Omnia Core container, run:

Run on: omnia_core container
pulp status | jq -r '.content_settings.content_origin'

Sample output: https://172.16.255.254:2225

Remove the https:// prefix and use only 172.16.255.254:2225.

Configure Compute Nodes to Pull from Pulp

On each Kubernetes compute node:

  1. Edit the CRI-O configuration file:

    Run on: compute node
    vi /etc/containers/registries.conf.d/crio.conf

  2. Append this configuration at the end of the file:

    [[registry]]
prefix = "100.10.0.76:3445"
location = "100.10.0.76:3445"

[[registry.mirror]]
location = "172.16.255.254:2225"

  3. Reload and restart CRI-O:

    Run on: compute node
    systemctl daemon-reload
systemctl restart crio

Pull the Image

Pull the image using the original registry reference (CRI-O transparently redirects to Pulp):

Run on: compute node
crictl pull 100.10.0.76:3445/library/nginx:1.25.2-alpine-slim

The image will be retrieved from the Pulp mirror automatically.

Next Steps