Updating a DPU backed ESXi using vLCM fails as the DPU doesn't get automatically restarted during the update, causing a version mismatch

Products

• VMware vSphere ESXi

Issue/Introduction

Manually shutting down the DPU during the first restart process allows ESXi to update correctly however then this issue described in KB 391443 occurs and the rollback fails

BMC communication errors after reboot present in /var/log/syslog.log:

Er(11) dpu-ctl[21###70]: libbmcal:Bmclib_Validate_Redfish_Credentials:746: Failed to Authenticate with BMC
Er(11) dpu-ctl[21###70]: libbmcal:Bmcalib_Init_x86_Dell:225: Failed to validate IPMI Bootstrap account credentials.
Er(11) dpu-ctl[21###70]: libbmcal:Bmcalib_Init_x86:1020: vendor-specific initializaiton of BMCAL failed
Er(11) dpu-ctl[21###70]: libbmcal:Bmcalib_Init:1121: BMCAL initialization failed!

BMC NIC vusb0 isn't present - seen either through UI or using esxcfg-nics -e command

BMC vSwitch and its associated vmknic is not present - seen either through UI or using esxcfg-vswitch -l command

Environment

ESXi 8.0 U3

Cause

BMCNetworkEnable advanced is set to disabled

For the DPU management, vSphere ESXi needs to have a connection BMC (to the iDRAC) - without this network, ESXi won't be able to get information about the DPU; in particular its state and to send a restart command 

Resolution

Ensure BMCNetworkEnable advanced setting is enabled.

To enable use below command:

esxcfg-advcfg -s 1 /Net/BMCNetworkEnable

Reboot the ESXi and ensure that the BMC vSwitch (named vSwitchBMC) is created with uplink vusb0. To check, use below command:

esxcfg-vswitch -l

Once confirming the above, reboot the ESXi once more to restart the DPU