In vSphere infrastructures, ESXi host and virtual machine high availability are as important to uptime as hardware availability, adding peace of mind with another layer of redundancy. VMware High Availability's range of settings allow users to fine-tune virtual availability thresholds.
Traditional system uptime focuses on hardware redundancy -- RAID for storage or fail-over functionality for entire servers -- and leaves out the operating system and application. Both hardware and software failures will crash ESXi hosts. Even if the host is working but will not respond to network traffic, it is useless for the virtual environment.
Virtual high availability comes from software features. For vSphere environments, VMware High Availability (HA) deals with ESXi host crashes as well as failure at the virtual machine's (VM) OS and application levels. While, in the past, VMware HA configuration and reliability were flaky, the vSphere 5 Suite seems to have made huge improvements on that front.
VMware uptime options
Take the FT and HA quiz
Recovery: SRM and Replication
VM availability is monitored by VMware Tools, installed on the guest VM. VMware HA receives heartbeats, or communication pings, from the VM at set intervals and at a set duration. If these thresholds aren't met, then HA will reset the VM. Application availability and failure detection also rely on a heartbeat, this time sent from an application to an agent on the host.
Because VMware HA restarts hosts and virtual machines after a failure, it requires a host cluster, defined as two or more ESXi hosts using shared storage and VMware Distributed Resource Scheduler. Be sure that DRS is enabled on any HA clusters. Both DRS and HA come in vSphere Standard, Enterprise and Enterprise Plus editions.
Choosing VMware HA settings
Adjust host and VM HA settings in accordance with your environment's uptime and performance requirements and your available resources. Edit HA settings in the vSphere Client by right-clicking on a cluster, then selecting Edit Settings and vSphere HA.
Enable Host Monitoring, one option, is recommended for use, except when doing host maintenance.
Admission Control and Admission Control Policy options allow the VMware admin to configure availability constraints and the policies that they will enforce.
Use Virtual Machine Options to configure how VMs restart and respond to host isolation. By default, VM configuration options are applied to all VMs in a cluster, but you can overwrite them on individual VMs when needed. VM Monitoring can be set at Low, Medium or High sensitivity monitoring of VM heartbeats or disabled entirely. Disable VM monitoring on a particular VM if you needed to take it offline or shut it down to isolate a suspected issue, such as an errant process that is affecting the entire network.
Use Datastore Heartbeating when the management network fails.