Skip to main content

📚 Making MTU 9000 Persistent on XCP-ng (OVS)

Making MTU 9000 Persistent on XCP-ng (OVS)

📖 Purpose

This guide explains how to correctly configure MTU 9000 on Open vSwitch (OVS) inside XCP-ng to ensure jumbo frames survive across reboots.

📖 Requirements

    XCP-ng 8.x or newer Open vSwitch in use (default in XCP-ng) SSH access to the hypervisor Root privileges

    📖 Background

      XCP-ng uses Open vSwitch (OVS) to manage VM networking. Setting MTU at the Linux NIC level is not enough; OVS controls the real MTU behavior. XCP-ng resets network bridges at boot, so MTU settings must be reapplied automatically.

      📖 Steps

      🛠 Step 1: Create the MTU Fix Bash Script

      sudo nano /usr/local/bin/fix-ovs-mtu.sh

      Paste this:

      #!/bin/bash
sleep 20
ovs-vsctl set interface eth4 mtu_request=9000
ovs-vsctl set interface xenbr4 mtu_request=9000
echo "$(date) - MTU 9000 applied to eth4 and xenbr4" >> /var/log/fix-ovs-mtu.log

      Make it executable:

      sudo chmod +x /usr/local/bin/fix-ovs-mtu.sh

      🛠 Step 2: Create the systemd Service

      sudo nano /etc/systemd/system/fix-ovs-mtu.service

      Paste this:

      [Unit]
Description=Fix OVS Interfaces MTU to 9000 After Boot
After=network-online.target openvswitch-switch.service

[Service]
ExecStart=/usr/local/bin/fix-ovs-mtu.sh
Type=oneshot
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target

      🛠 Step 3: Enable the Service

      sudo systemctl daemon-reload
sudo systemctl enable fix-ovs-mtu.service
sudo systemctl start fix-ovs-mtu.service
sudo systemctl status fix-ovs-mtu.service

      🛠 Step 4: Validate

        Check service status: 
        sudo systemctl status fix-ovs-mtu.service
          Check MTU on physical NIC and bridge:
            Verify inside OVS: 
            ovs-vsctl list interface eth4 | grep mtu
ovs-vsctl list interface xenbr4 | grep mtu

            📖 Expected Results

              eth4 MTU = 9000 xenbr4 MTU = 9000 VMs' VIFs inherit MTU 9000 Jumbo frames supported end-to-end

              📖 Troubleshooting

              Symptom Likely Cause Solution MTU resets to 1500 after reboot Systemd service not enabled Check service status Ping fragmentation on large packets Switch ports not configured for MTU 9000 Enable jumbo frames on switch ports iperf3 speed low MTU not properly set at all layers Double-check NIC, bridge, VM, and switch

              📖 Bonus: Quick MTU Validation Script

              #!/bin/bash
echo "Checking eth4:"
ip link show eth4 | grep mtu
echo "Checking xenbr4:"
ip link show xenbr4 | grep mtu
echo "Checking VMs VIFs:"
ip link | grep vif | grep mtu

              📖 Final Notes

                Always verify MTU end-to-end: Storage → Switch → Hypervisor → VM. Setting mtu_request in OVS and using a small systemd service guarantees success across reboots. Jumbo frames dramatically improve 10GbE performance, especially with storage traffic like iSCSI and NFS.

                🎯 End of Tutorial

                Back to top