While testing the recent UCX PR I noticed I was getting this warning:

--------------------------------------------------------------------------
WARNING: There was an error initializing an OpenFabrics device.

  Local host:   cn-priv-01
  Local device: hfi1_0
--------------------------------------------------------------------------
[cn-priv-01:3767216] select: init of component openib returned failure

The problem is, the ipoib interface is working fine on the nodes in this run - 
and there's no more information about what the error might have been. Can 
anyone shed any light on why this might be happening? I do not see this with 
OMPI 4.0.3.

---
Michael Heinz
Fabric Software Engineer, Cornelis Networks

Reply via email to