Issue
On NVIDIA QM9700 InfiniBand switches, some ports may appear down and show an
error such as:
This issue is often caused by outdated CPLD firmware and can be resolved by
updating the CPLD version on the switch.
Root Cause
The EEPROM error is typically due to compatibility or communication issues
between the switch and the optical cables/transceivers. A CPLD firmware update
resolves low-level hardware interface bugs related to EEPROM communication.
Solution: Update CPLD
Firmware
Requirements
· Linux server with SSH access to the switch
· updateswitchcpld_v3.1.tgz
CPLD update tool
· Switch IP (management or IPoIB)
· Admin username and password
Step-by-Step Procedure
1. Transfer and Extract the
Update Tool
Copy updateswitchcpld_v3.1.tgz
to your Linux host having connectivity to ib switch and extract it:
tar xzf updateswitchcpld_v3.1.tgz cd updateswitchcpld_v3.1
Use the following command to start the CPLD update process:
./updateswitchcpld --managed -t <SWITCH_IP> -u <ADMIN_USERNAME> -p <PASSWORD> --os mlnx-os
Example:
./updateswitchcpld --managed -t 10.152.15.55 -u admin -p <password> --os mlnx-os
· The switch will reboot during the CPLD update.
· Save configuration before
running this step.
3. OS Version Compatibility
Note
There is a known compatibility issue with MLNX-OS version 3.11.4002
when running the CPLD update script. If you are using this version:
· Upgrade temporarily to a compatible version like
3.12.1000 or 3.11.4000
· Run the CPLD update
· Revert back if needed
Switch OS Upgrade/Downgrade
Commands
show images image boot next <image_name> configuration write reload
4. Confirm CPLD Upgrade
After reboot, validate the CPLD version:updateswitchcpld --managed -t <SWITCH_IP> -u <ADMIN_USERNAME> -p <PASSWORD> --os mlnx-os --verbose --check_cpld
Post-Update
· Ensure ports previously showing “EEPROM” errors
are now active.
· Validate the cables using ibdiagnet
or the switch web UI/CLI.