Attendees:
Bob Monkman (Intel)
Alexander Vul (Intel)
Martin Halstead (HPE)
Prodip Sen (HPE)
Rafal Sztejna (Intel)
John Leung (Intel)
Joseprabu Inbaraj (AMI)
...other please add your names
Agenda:
Walk on item- Lab environment
DMTF Redfish Overview John Leung (Intel)
Plug In Open Discussion
Minutes:
Captured by BobM...please correct, refine, add where needed...I may have misinterpreted some items
Topic: Lab Environment (Alex Vul leads)
- We need to think about a shared lab setup where we can test code on sets of servers
- Alex asked if HPE could host a pod for this purpose and Martin H took an AR to look into it
- AR : Alex will check into Intel lab facilities as well
- Alex noted we should also look into formulating a release schedule
Topic DMTF Redfish Overview
- John Leung is the Intel rep for DMTF Redfish and agreed to give an overview next week's TSC meeting (20 Minutes)
- For now, we discussed that Intel with partners had compiled an initial list of proposed extensions/additions to Redfish specs
- It was requested that they be posted on the wiki for review and discussion- AR Alex V to post a strawman.
Topic: Plug In
Alex V indicated that a Rack Manager Plug In is needed
There had been a contribution to OCP some time ago - this was/is referred to or became OpenRMM. Can we just leverage that?
What would be the implications for the standalone OpenRMM project if we did that?
Conversation turned to the fact that there are many "unmanaged" racks out there. Would such a rack/resource need to interface to a plug-in if there is not an explicit rack manager associated ?
John Leung noted that he is also an Intel rep for OpenRMM (or was it Open BMC?- my notes are not clear)
John noted that from an OCP standpoint ensuring that a bottoms up mgmt of open firmware, OCP board level mgmt does bubble up to Rack Level mgmt interface.
There is a strawman proposal in ODIM wiki to have a way to get info on servers and other resources within a rack. We do need something fleshed out here.
Martin noted that one does have multi-vendor server, switch , storage etc within racks out there. It would be good to have a proposal to create a model for managing these resources at a rack level.
Redfish has a notion of chassis type= rack, we need this in the plug-in concept/model, even if some racks do not have a formal Rack Mgr, ODIM Resource Aggregator could/should propose to have a plug-in to ensure we can discover, inventory, etc resources within the chassis , within a grouping of racks.
Alex noted that we do have a discovery method (did not capture where) and martin noted that Redfish has no formal way to create a topology of physical heterogeneous resources, and so we need to flesh something out so that we can discover these.
We need a way to advertise/discover/inventory these resources as they come on line.
We need a NB client interface to create/if manually at first, an inventory of added devices. Ultimately the rack itself could eventually managed in an automated way if there is a plug-in that can generate events and logic to process said events.
After a good discussion, there seemed a good understanding of the need for a Plug In architecture here and a proposal is/can be posted on the wiki for review and further discussion.
End of meeting.