Open ricardoas opened 6 years ago
I've tried my best to put it to work but with our system but I failed. I can, if you like, compile and configure all servers, the documentation is quite straight forward on this. On pod-manager I did 50% of the configuration on a ansible playbook for easy replication. On the others modules I didn't do anything. I started the work thinking that after putting it together at some place I would point to out system and it would discover the enclosure SH. But it's not how it works. In a nut shell: Pod-manager communicates with the other modules (RMM, LIU and PSME and RSS) RMM is a module that collects info on a Rack, so it would get information about power, colling, etc. I can't tell exactly how it works because it's user guide only tell you how to compile it. LIU is a Linux Image Utility: it's used by pod manager to deeper discover of hardware. Basically it boots the server via iPXE and starts a system via iscsi. After the boot the image inspect the hardware and feeds it back to pod-manager. PSME server is a rest API server that allow agents to inform the PSME about the hardware they are installed. So it's seems to me that you need the specific agent installed on each server to collect information. RSS is a storage service which I didn't look at the documentation. But based on the information I gathered from others modules it's agent-based too. So it seems to me that the system was conceived as a REST service API that would expect agents to feed them with information about the hardware they are supposed to administer. Not the other way around (the system goes to the server and discover what it has using Redfish). The only thing I could notice regarding redfish in this system it that they provide a redfish API on each developers of pod manager software could use then as a way to discover (using their agent -rest server schema) and from that it would discover and compose logical hardware. It's my understanding that we expected this system to be such software and that it would get information from our redfish-server and provide an interface to use it. It seems to me that this IntelRSD system is something like our system, that collects info from a hardware and provides it's information via Redfish interface.
Managed to compile and run PSME server with compute-agent. Please see the conversation in the #general channel of slack for update. I have opened a issue in the intelRSD project expecting some guidance on how to proceed. I didn't create a playbook for this part (PSME) and asked in the slack if anyone wants it. This is blocked until some response from intel RSD team or someone give me new directions or PO closes this issue if it's not relevant anymore.
intelRSD tool only talks to hardware using IPMI and not redfish. It only uses redfish to talk between the modules (PODM with PSME, RMM, etc) and to serve the final user. It's not possible to use this with our project. This can be closed. Waiting PO review and closure though.
Information available at: https://github.com/01org/intelRSD