Sundar Yamunachari wrote: > > The design document for the AI spring release is at > http://www.opensolaris.org/os/project/caiman/auto_install/Design_doc_delta_for_AI_Spring_2009.pdf. >
Sundar, Here are some questions/comments based on my initial pass through the document: Section 1.1 The text says "We need to create a SMF service for the Apache server..." and then goes on to say "The AI web server and service registration process should be controlled by the SMF installadm service." Based on this it sounds like the Apache server *will* become an SMF service while the AI web server *might* be controlled under SMF. Is this interpretation correct or is the plan to have both of them under SMF for this release? Just to be certain, the final paragraph of this section is specifying what the ideal solution would be but is saying that this will *not* happen for the spring release. Correct? Section 1.2 Several uses of "could" in the paragraph beginning with "If DNS configuration is not possible..." Can you tighten this down to say what you intend to do for the spring release? For the DNS-naive like me, what are the implications of using DNS unicast? I believe I understand this means that an install client can reside on a different subnet than the server, but specifically which server functions (DHCP, DNS, TFTP, http) could be on a different subnet and which ones would have to be on the same subnet as the client (based on what little I know I'd assume DHCP at least would need to be on the local subnet while http can be anywhere). Section 1.3 I like this section because it unambiguously states what you're planning and doesn't say 'could' or 'should'. :-) Section 1.4 Also makes statements I like: "...the format of the composite manifest will be changed as follows..." Section 1.5 The 'installadm list' subcommand should be a big help in allowing the tests to verify that a previous command completed successfully (giving a method to list out the services). Could there be an option to list just the last set of services that were created/started (e.g. after starting a service a test could then use this command to retrieve the name of the service just started vs. any others that might have been running already)? Section 1.6 Does testing this out just mean that the tests should create manifests with both forms of the MAC ID (with and without colons) and make sure that both work? Regards, Andre
