Using the TeamPolicyUpDelay setting

Products

• VMware vSphere ESXi

Issue/Introduction

• Link failure detection and teaming/failover policies allow ESXi to redirect traffic to working uplinks when one or more network interfaces on the host become disconnected.
  
  
• The uplink is re-enabled once the connection is restored, based on the "Link Status Only" failure detection policy.
  
  
• In an active/standby uplink configuration, if failback is enabled, traffic is moved back onto the link. By default, there is a 100 ms delay before this occurs.
  
  
• In an active/active uplink configuration, it will, by default, fail back the traffic based on the load balancing policy. 

 

Environment

VMware vSphere ESXi

Cause

The network card may report that the link state is active well before full connectivity is restored. For example, this can be due to Spanning Tree Protocol (STP) operation, vPC recovery delays, etc., which can delay connectivity for 30+ seconds.

Resolution

• The setting Net.TeamPolicyUpDelay can be changed in Advanced System Settings to adjust the delay in milliseconds before an uplink is re-enabled. TeamPolicyUpDelay defines the time, in milliseconds, that the ESX host waits after detecting that a pNIC has transitioned from a down state to an up state before the host starts using the pNIC again for network traffic. This delay helps to ensure that the pNIC has fully recovered and stabilized before it is put back into use. Usually TeamPolicyUpDelay is modified when physical switch upgrades are happening. By bringing an uplink to UP state immediately, it can impact the traffic as the connections may not reset immediately via the new uplink.
  
  
• For example, a setting of 60000 will wait 60 seconds (1 minute) before a network interface is considered usable.
  
  
• This setting takes effect immediately in ESXi 8.0.

Additional Information

The Net.TeamPolicyUpDelay is a global advanced system setting on the ESXi host that defines the wait time (in milliseconds) after a physical NIC link state is "Up" before the host considers the link active for teaming and load balancing.

Please also refer to the Broadcom Knowledge Base (KB) articles :

Duplicate packets are sent out of the ESXi to physical switches when both the available uplinks on a DVS flap causing traffic outage

Wait period defined by TeamPolicyUpDelay not honored resulting in network outage.

Changing the network teaming failback delay fails in ESXi Host

TeamPolicyUpDelay is not working as expected when beacon probing is enabled on a DVPG in DVS.

Changing the network teaming failback delay (TeamPolicyUpDelay) is not working as expected on the ESXi hosts when the VDS has DVPGs with Beacon Probing enabled.