In Linux networking, logical interfaces (bond, vlan, bridge) operate at different layers of the network stack than physical interfaces. The MTU (Maximum Transmission Unit) setting follows a hierarchical relationship:

Physical NICs (eth0,eth1) → Bond interface → VLAN interfaces → Bridge interfaces

The effective MTU is determined by the lowest value in the chain. Consider this practical example:

# Physical interfaces
ip link set eth0 mtu 9000
ip link set eth1 mtu 9000

# Bond interface inherits from physical
ip link set bond0 mtu 9000

# VLAN interface (must be ≤ bond0 MTU)
ip link set vlan10 mtu 1500

# Bridge interface (must be ≤ vlan MTU)
ip link set br10 mtu 1500

When configuring MTUs in complex network stacks:

• Jumbo frames (MTU > 1500) require consistent settings across all layers
• Bridge interfaces cannot have MTU larger than their member interfaces
• VLAN tagging adds 4 bytes overhead (account for this in MTU calculations)

To verify your MTU settings are properly applied:

# Check current MTU for all interfaces
ip -br link show | awk '{print $1,$4}'

# Test end-to-end MTU (from domU through all layers)
ping -M do -s 8972 10.0.0.1  # (1500 MTU: 8972 = 1500 - 28 ICMP header)

Incorrect MTU settings can cause:

• Packet fragmentation (when higher MTU interfaces send to lower MTU interfaces)
• TCP blackhole connections (when PMTUD fails)
• Reduced throughput (when not using optimal jumbo frames)

For your specific bond+vlan+bridge setup:

# Set consistent MTU across all layers
for intf in eth0 eth1 bond0 vlan10 vlan20 vlan30 br10 br20 br30; do
    ip link set dev $intf mtu 9000  # Or your desired MTU
done

# Verify with
for intf in eth0 eth1 bond0 vlan10 br10; do
    echo -n "$intf: "; cat /sys/class/net/$intf/mtu
done

When encountering MTU-related issues:

# Check for silently dropped packets
ethtool -S eth0 | grep -i 'drop\|error'

# Monitor interface statistics
ip -s link show bond0

# Check kernel messages for MTU warnings
dmesg | grep -i mtu

In complex network configurations combining bonding, VLANs, and bridges, the MTU relationship between logical and physical interfaces becomes crucial for optimal performance. The typical flow looks like this:

Physical interfaces (eth0, eth1) 
  → Bond interface (bond0) 
    → VLAN interfaces (vlan10, vlan20, vlan30) 
      → Bridge interfaces (br10, br20, br30)

Here's what actually happens with MTU settings in this stack:

• Physical interfaces have their own MTU (typically 1500 by default)
• Bond interfaces inherit the lowest MTU of member interfaces
• VLAN interfaces inherit from their parent interface
• Bridge interfaces use the MTU of their first added interface

To ensure proper MTU handling:

# Example: Setting Jumbo frames consistently
# Physical interfaces first
ip link set dev eth0 mtu 9000
ip link set dev eth1 mtu 9000

# Then bond interface (auto inherits)
ip link set dev bond0 up

# VLAN interfaces (explicit setting recommended)
ip link set dev vlan10 mtu 9000

# Bridge interfaces
ip link set dev br10 mtu 9000

Common symptoms and solutions:

For Xen domU configurations:

# In dom0 configuration (xenbr0 example)
auto xenbr0
iface xenbr0 inet manual
    bridge_ports vlan100
    bridge_stp off
    bridge_fd 0
    mtu 9000
    
# Corresponding domU config
vif = ['mac=00:16:3e:xx:xx:xx,bridge=xenbr0,mtu=9000']

Remember that virtual interfaces in domUs cannot exceed the MTU of their parent bridge in dom0.

• VLAN tags consume 4 bytes of MTU space
• Some NICs have hardware MTU limits
• Routing equipment must support end-to-end MTU
• Always test with 'ping -M do -s' to verify path MTU