Fix iSCSI TCP Retransmits with Force10 Switches, EqualLogic, and VMware

I was seeing high TCP retransmit rates on an iSCSI SAN built on Dell Force10 switches, EqualLogic storage arrays, and VMware ESXi hosts. The fix was configuring every switch port that connects to an EqualLogic array or a vHost iSCSI NIC with jumbo frames, flow control, and RSTP edge-port settings.

Force10 iSCSI Port Configuration

config
protocol spanning-tree rstp
no disable
interface GigabitEthernet x/x
description EQL or vHost iSCSI Port
no ip address
mtu 9252
switchport
flowcontrol rx on tx on
spanning-tree rstp edge-port
no shutdown
!

What Each Setting Does

  • mtu 9252: enables jumbo frames on the port for iSCSI traffic.
  • flowcontrol rx on tx on: lets the switch and storage pause traffic instead of dropping packets.
  • spanning-tree rstp edge-port: brings the port up immediately without waiting on spanning-tree.

With these settings in place, TCP retransmits dropped below 0.05%.