Hi Folks I noticed that the NERSC (carver/edison) MTT smoke tests are failing now. I also see a lot of ivy cluster runs are also failing. All the nersc runs are failing with:
c1479:05071] OPAL ERROR: Bad parameter in file util/attr.c at line 431 [c1479:05071] [[57033,0],0] ORTE_ERROR_LOG: Bad parameter in file util/attr.c at line 57 [c1479:05071] Signal: Segmentation fault (11) [c1479:05071] Signal code: Address not mapped (1) [c1479:05071] *** End of error message *** the mpirun command line is mpirun --bind-to none -np 32 --mca coll ^ml --mca btl self,vader,openib --prefix /global/u2/h/hpp/mtt_carver_tmp/installs/8v68/install ./c_hello Before people begin blaming this as a cray thing, this is from the NERSC carver system which is an ibm dataplex system running redhat and using MLNX connectX HCAs. Anyone else seeing these failures? Howard