A network using AI is composed of the following components.
A DHCP server that provides the AI client with host information.
One or more IPS (Image Packaging System) repositories that provide the software packages to install on the AI client. The system image for the AI client may also be created from an archive located on the network.
An AI server that has configuration instructions for the AI clients.
One or more AI clients.
Figure 2 AI Network Example
The DHCP server, IPS server, and AI server do not need to be hosted on separate systems. In particular, installing the AI server and the DHCP server on the same system eases administration steps because the installadm command will update the DHCP service if the DHCP service is co-located with the install service.
The AI server itself can contain the following components:
One or more install services. Each service is configured to match the architecture and OS to be installed on the AI client.
One or more AI manifests. An AI manifest provides installation instructions, such as the disk layout to use or the packages to add.
Optional system configuration profiles. These profiles provide system configuration information, such as which timezone or name service to use.
In addition, you can create an IPS package to deliver a first-boot script to the AI client during the installation process. This script can perform additional installation or configuration steps that cannot be done using an AI manifest or system configuration profile, such as adding a third-party utility to an AI client.
The DHCP server manages network connections between the AI server and the clients. It provides necessary network information as well as the AI server location to the clients. You should co-locate the AI server and DHCP server to facilitate administration.
A SPARC client can be configured to locate the AI server without DHCP, but an x86 client cannot. For alternative ways to perform AI without DHCP, see Installing an x86 AI Client and Installing a SPARC AI Client.
The client stems you want to install must be able to access an Oracle Solaris Image Packaging System (IPS) software package repository or an Oracle Solaris Unified Archive. A repository is a location from where software packages are retrieved. The location is specified by a Universal Resource Identifier (URI). The IPS package repository can be on the AI server, on another stem on the local network, or elsewhere on the Internet. See Configuring Publishers in Adding and Updating Software in Oracle Solaris 11.3 for information about accessing a package repository. The IPS server can also provide any first-boot scripts that are needed to completely configure the AI client.
To use AI to stall systems over the network, you must first set up an AI install service on an AI server. For the complete procedure, see Setting Up an AI Server. Part of the procedure shows how to create a static network address for the AI server because the IP address for the server is included in the files created for each AI client. If the IP address of the AI server changes, then the configuration files for all AI clients have to be re-created.
Each server can include one or more install services. You must create an install service for each version of the OS and for each client architecture you need to support. For example you could have an install service for SPARC clients booting Oracle Solaris 11.2, another for SPARC clients booting Oracle Solaris 11.3, and then two more to provide the same services for x86 clients. Each install service includes a SPARC or x86 boot image, one or more installation instruction files (AI manifests), and optional system configuration profiles. Creating an Install Service provides instructions for creating and maintaining install services.
The boot image provided by the AI server is not a complete installation. The boot image creates a configuration on the stem in which the installation can run. client stems must access an IPS package repository or an archive to complete their installations.
An AI manifest includes client provisioning or installation instructions. Each AI client uses only one AI manifest, although many AI clients can share one manifest. The AI manifest specifies one or more IPS package repositories where the AI client retrieves the packages needed to complete the installation. You can use an archive can be used in place of the IPS packages. The AI manifest also can include the names of additional packages to install, and information such as the target installation device and partition information. If you need to install two client stems with the same version of the Oracle Solaris OS but they need to be installed differently in other ways, then create two AI manifests associated with one AI install service. The different AI manifests can specify different packages to install or a different slice as the install target, for example. See Defining AI Client Installation Parameters for information about creating and customizing AI manifests, either prior to booting the AI client or dynamically at installation time.
If AI clients need to have different configurations applied, then create multiple system configuration profiles for the install service. The different system configuration profiles can specify a different network or locale setup or a unique host name and IP address, for example. A profile that sets up the time zone may be used by several AI clients. See Defining AI Client System Configuration Parameters for information about profiles.
If no profiles are configured for an AI client, an interactive tool will prompt for system configuration information after the AI client boots once the installation is done. You would need to manually provide the information as prompted.
To include configuration that cannot be expressed in an AI manifest or system configuration profile, you can include a script that runs at first boot. See Running a Custom Script During First Boot for detailed information.
The installation begins when you boot the AI client. When the AI client boots, the client is directed to the AI server, and the client accesses the correct install service and the correct AI manifest and system configuration profiles associated with that service. Assigning Customizations to AI Clients explains how an AI server identifies the correct AI manifest and system configuration profiles to use when an AI client is installed.