Troubleshooting#
Interface name mismatch during provisioning
During DGX node provisioning, an error appears:
This node (``dgx-01``) was booted using an interface (``ens8f1np1``) which is not defined in the node object
Cause: RHEL and Ubuntu resolve network interface names differently on DGX B200 hardware. Ubuntu uses PCIe path-based naming (
enp170s0f1np1,enp41s0f1np1); RHEL uses slot-based naming (ens8f1np1,ens2f1np1) for the same physical interfaces. Both names are recognized by the kernel. This is a naming policy difference between the two distributions, not a hardware or cabling issue. The node object in BCM is defined with the PCIe path names, which do not match what RHEL uses during provisioning. To confirm the kernel-detected names, check/var/log/node-installerlogs.Resolution: Update the interface names in cmsh to match what the kernel detected:
# At the OS/shell level on the BCM head node, run cmsh [clio-bcm11-head01]% device [clio-bcm11-head01->device]% use dgx-01 [clio-bcm11-head01->device[dgx-01]]% interfaces [clio-bcm11-head01->device[dgx-01]->interfaces]% use bond0 [clio-bcm11-head01->device[dgx-01]->interfaces[bond0]]% removefrom interfaces enp170s0f1np1 enp41s0f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*]% use enp170s0f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[enp170s0f1np1]]% set networkdevicename ens8f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[ens8f1np1*]]% set alternativehostname enp170s0f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[ens8f1np1*]]% use enp41s0f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[enp41s0f1np1]]% set networkdevicename ens2f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[ens2f1np1*]]% set alternativehostname enp41s0f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[ens2f1np1*]]% use bond0 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[bond0*]]% set interfaces ens8f1np1 ens2f1np1 [clio-bcm11-head01->device*[dgx-01*]->interfaces*[bond0*]]% commit