[Ofmfwg] Sunfish meeting notes for May 4, 2026

[Lee, Peter]

  *   Datacenter use case
     *   Continue discussion about rabbits. Mike found more details about rabbit architecture. The rabbit node set everything up, including setting up the PCIe switches to connect between the compute nodes and the NVMe drives. Need to determine whether the existing NVMeoF design in mind can work with the rabbit architecture. Further discussions are needed to dive into the details to determine.
  *   Code changes
     *   Russ continues working on the eventing support but ran into 2 issues while testing.
        *   The Redfish architecture allows you to subscribe for events on any objects. With the current code, a Sunfish client subscriber may run into event loops if it subscribes for objects it owns. The objects that cause these loops are aggregation sources like Subscription. Russ proposes that we don't add additional checks into the code to prevent this. Instead, we document in the specification that if the Sunfish client subscriber does not want event loop, it needs to exclude those resource types in the event subscription.
        *   In the current delete object code, there is an order of execution issue. When object is deleted, it gets deleted from the database. But when we then go to create the event notification, the object is no longer available in the database for us to retrieve the object details to put in the event notification. Russ will fix this issue.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ofmfwg/attachments/20260504/761d82ae/attachment.htm>