Dell server dual DPU issue related to Add VI / cluster in loading switch configuration page.

Products

• VMware Cloud Foundation

Issue/Introduction

Symptoms:

During switch configuration phase of Add VI workload or Add secondary cluster workflow using Dell servers that have dual DPUs and also a vusb0 interface the below error may be encountered.

Presence of the vusb0 interface can make the total interface list an odd number (7 in this above illustration), See below in hosts table to confirm the presence of vusb0 interface in the host interface list.

 

Environment

VMware Cloud Foundation 5.2

Cause

There is a product issue due to which the algorithm to show available profiles in switch configuration page is unable to show any of the profiles and custom profile does not function properly when vusb0 interface is present and the total interface count is an odd number. This is not the typical case because most NICs have 2 interfaces thereby making the total interface count an even number.

Resolution

Currently there is no resolution.

Workaround:

Update the SDDC manager inventory to remove the vusb0 interface from the relevant table so that the product logic can display the switch configuration profiles accurately. This has 0 impact on the functionality of SDDC manager since this interface (vusb0) is a special interface and should not be used by SDDC manager for any of its configurations (uplink interfaces)

1. Take backup of the SDDC manager using SDDC manager or vsphere client (mandatory precaution to revert to this snapshot if any error gets inadvertently introduced with the following steps)
2. Login to sddc manager as user vcf using SSH and go to superuser mode (su)
3. login to sddc manager inventory with command "psql -U postgres -h localhost" and then in the sql prompt run commands "\c platform" (connect to platform db) and then "\x" (expanded output)
   
   
4. In the sql prompt for each of the hosts that you intend to use in the new workload domain / cluster perform the following operations:

select host_attributes from host where hostname='<host-fqdn>';

This will display the entire JSON string that represents the entire host attributes for this host. You now need to copy that entire string (including the opening and close angle brackets) into any temporary file.

Please make sure that upon copying this content into the file, the contents is a valid json (this is to ensure that the entire string was copied correctly with the braces), also make sure that the string is one line (no newlines present) and no extra spaces got inserted in the string value.

Now, edit the string to delete the last vusb0 interface (highlighted portion below) including the ',' (comma) after the entry preceding the vusb0 specific entry.

With the exception of the vusb0 interface, the contents will appear similar to the below output similar to below.

Now, verify that the modified string (JSON) also is a valid JSON

Copy the entire string and use it in command "update host set host_attributes='<modified-json-string>' where hostname='<host-fqdn>';" in the sql prompt of SDDC manager opened earlier.

The output of this command should state "UPDATE 1" without any error.

Now, reload sddc manager UI and in hosts view the interfaces for this host should show all interfaces except the "vusb0" interface (see below)

Once done for all the hosts, launch Add VI workload / Add cluster workflow and the issue should no longer be present in switch configuration page.