Don't make this mistake when upgrading to 10Gbit network

YouTube Video

Key points from the YouTube transcript:

• Problem: The YouTuber experienced significantly reduced upload speeds (around 6 Mbps) on a Proxmox cluster with Intel X710 10GbE network cards, despite having a 10Gb link. This issue only affected one of two identical clusters.

• Troubleshooting Steps Taken:

  • Firmware/BIOS Updates: Upgraded and then rolled back firmware on the affected servers. No change.
  • Reboots: Rebooted servers and the switch stack. No change.
  • Driver Investigation: Focused on the i40e kernel module for the Intel X710 NICs. Suspected driver issues.
  • Intel Driver Installation: Attempted to install the latest Intel i40e drivers from the official repository. This initially worsened the problem, resulting in complete network loss, before partially resolving the issue temporarily.
  • Kernel Module Rollback: Rolled back to the default i40e kernel module. The issue persisted, although briefly achieving normal speeds.

• Current Status: The problem remains unresolved. Upload speeds are stuck at around 6 Mbps.

• Next Steps: The YouTuber plans to:

  • Test with a different operating system and kernel version.
  • Test with a different switch.
  • Seek support on the Proxmox forums.

• Suspicions: The YouTuber suspects a driver issue, possibly a race condition within the i40e kernel module or a hardware problem affecting only some of the NICs (based on a forum thread they read). They also note the problem’s unusual and seemingly random nature, affecting only some servers in an otherwise identical setup.