Booting ESXi host without starting the vSphere HA (FDM agent) on startup

Products

• VMware vCenter Server

Issue/Introduction

This article provides how to prevent failover of virtual machine(s) during recovery of an HA enabled Cluster from a Datacenter wise outage.

Environment

VMware vCenter Server 7.0.x

Resolution

Please use it as per recommended steps given below in this KB article. VMware recommends that you use this KB under the direction of VMware Technical Support.
 
Recommended Steps

1. Boot All ESXi host of cluster using boot param "fdm=no".
2. Power on the vCenter Server VM. If the vCenter Server VM is deployed on the same cluster, It need to power on using the vim-cmd command. For more information, see Powering on a virtual machine from the command line when the host cannot be managed using vSphere Client (1038043).
3. Disable High Availability (HA) in vCenter Server UI.
4. Once HA is disabled, reboot all the ESXi Host normally without the boot param "fdm=no").
5. Once completed and HA protection is disabled on the cluster, the virtual machine(s) can be powered on.

Limitations

• Do not perform any steps other than what is specified under point 2 & 3 on the vCenter Server or ESXi host when host are booted with param “fdm=no”
• Any other step other than above may lead to unspecified behavior.
• All ESXi host(s) must be down prior to performing the above steps. Performing the above steps on a partially down cluster may lead to undesired behavior.

Prerequisite

• The vCenter Server should be greater than version 7.0 Update 1.
• The corresponding fdm agent on each ESXi should be upgraded (upgrade of fdm agent happens automatically whenever the ESXi host syncs to the upgraded vCenter Server on an HA enabled cluster). Verify this by running the “esxcli software vib list | grep fdm” command on the ESXi host.

Note:

1. Boot param fdm can be given to ESXi while booting through:
   
   a. Network (PXE boot).
   b. ILO.
   c. Physically on grub prompt.
    
2. The boot param “fdm” can have these values:
   
   a. fdm=no is used on boot time fdm agent service will not start after ESXi is booted up.
   b. fdm=yes is used on boot time is equivalent to boot ESXi without using fdm boot param. in this case ESXi will boot normally.