[Ofmfwg] Sunfish meeting notes for March 28, 2025

[Lee, Peter]

  1.  H3 CXL work
     *   Welly is working on some of the broken links.
     *   Russ was working on the H3 friendly agent but ran into an issue where there is information in the aggregation source object on Sunfish server that should not be there. This occurs after the agent has register with Sunfish and upload is completed. Not sure how the information got there but suspects it may be due to Python's equal operator behavior. Need to do more detailed debugging.
  2.  Documentation
     *   Will put implementation related text into the appendix. This includes the flow diagrams we have worked on. Later, if we want to split the appendix into a separate document, we can do that.
     *   Need to add the work Russ did on OFMF server. Russ also wants to add some additional details to the registration process as, right now, it is kind of vague in some of the places.
     *   Want to get the documentation updated by end of April to get to v0.5.
     *   If any issues with documentation, send to Phil and create a new GitHub issue.
  3.  Github repository consolidation
     *   Review the consolidation discussion from last week.
     *   Will defer to Russ but will leave things as is, so Russ can complete his work.
  4.  Flux resources from Broker 0 to the compute brokers
     *   Mapping of the closest CDI resources to the computes
        *   Mike have not yet figure out how this work but will know more next week once he gets a chance to work more on this.
  5.  Flux Prolog/Epilog integration work, burst buffer deployment
     *   Mike is working on this now. Know the steps and just need to be put it together. From Flux viewpoint, resources from Sunfish will be active when in use and available when not in use.
     *   Mike will work to complete the burst buffer deployment and get it to Livermore when completed.
  6.  NVMeoF Sunfish Agent
     *   There was a question of how NVMeof devices show up in the topology for lstopo. When a host connects to NVMeof, the NVMeof device appears as local NVMe device. No one on the host will know the device is remote except the NVMe driver.
     *   Mike shared the following diagram. There were discussions that the diagram may not be accurate. Mike will look through the resources directory for the mockup on OFMF server to make the necessary changes to the diagram.
        *   [cid:image001.png at 01DB9FB7.D35360D0]
  7.  Meeting logistics for next week
     *   Mike is traveling next Monday. Will let people know by Sunday if we will meet on Monday.
     *   No meeting next Wednesday.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ofmfwg/attachments/20250328/ef0195f3/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 218269 bytes
Desc: image001.png
URL: <http://lists.openfabrics.org/pipermail/ofmfwg/attachments/20250328/ef0195f3/attachment-0001.png>