VIA images the rack components in a pre-determined order, which is determined by the availability of network route to the different components of the rack. All switches are imaged first. This enables VIA to access the servers through the switches for imaging. The imaging order is as follows.
The management switch is the main access gateway through which the Cloud Foundation management data is routed. The management ports of the ToR switches, inter-rack switches, and the physical servers are connected to the management switch. The data ports of the ToR switches are also connected to the management switch. This enables VIA to communicate with the servers over both management and data network through the management switch. VIA is also connected to the rack through a designated port on the management switch. It is therefore required that the management switch is the first component imaged by VIA in order to obtain access to the other components of the rack. VIA currently uses an IPMI connection to image the management switch.
ToR switches and Inter-rack switches (inter-rack switches are on rack 2 only)
Inter-rack and ToR switches are imaged in parallel.
Inter-rack switches inter-connect multiple racks enabling a scale out architecture for the datacenter. They create an stretched L2 backplane between racks.
ToR switches provide connectivity to servers in each rack out to inter-rack switches. The first pair of ToR switches provide connectivity to your datacenter network.
The management ports on the servers become accessible to the management switch during the course of imaging/configuration, which in turn make the management ports accessible to VIA through the management switch. Once all the switches are imaged and configured, the data ports of the servers become accessible to VIA through the ToR switches, which then proceeds to image the servers in parallel.
For each component that is being imaged, the following tasks are performed.
Rack components are discovered using the DHCP service running on the VIA VM. The DHCP Service uses the device type information to identify the device being discovered. Apart from the device type information, the DHCP service also uses hardware vendor specific strings to determine whether the switch being imaged is a management, ToR, or inter-rack switch.
The first component to be discovered is the management switch. The DHCP service hands out a pre-determined IP address for the management switch followed by a PXE image specific to the management switch.
After the management switch is imaged, the ToR and inter-rack switches are discovered and imaged. The management switch also discovers the IPMI network of the servers. This allows VIA to initiate imaging of the servers. The ToR switch enables discovery of the data network of the servers which is used to receive the installation image delivered by the DHCP service.
Image installation refers to installing software on the components to make them operational. The software depends on the component type - an Operating System for switches and a Hypervisor for servers.
This step in the imaging process ensures that the components of the rack work like a homogenous system. Configuration of each rack component is different. If any configuration step fails for the management, ToR, or inter-rack switches, imaging stops at that point and cannot proceed. If a configuration step fails for the server, imaging for that server cannot be completed but the remaining servers in the rack can be imaged.
Imaging is a multi-step process.