Cluster settings
Before the addition of attached storage, the VMware vCenter Server® solution didn't enable advanced features such as VMware® vSphere Distributed Resource Scheduler (DRS) and VMware vSphere® high availability (HA). With the addition of the NFS attached storage device, these features are enabled on the cluster with the settings that are listed in the following sections.
vSphere Distributed Resource Scheduler
Two main features are enabled when you turn on vSphere DRS on a cluster: load balancing and power management.
Load balancing
With load balancing, the distribution and usage of CPU and memory resources for all hosts and virtual machines (VMs) in the cluster are continuously monitored. DRS compares these metrics to an ideal resource usage given the attributes of the cluster’s resource pools and VMs, and the current demand. It then completes or recommends VM migrations as needed.
When a VM is first powered on in the cluster, DRS tries to maintain proper load balancing by either placing the VM on an appropriate host or by making a recommendation. The placement or recommendation settings are set in the DRS Automation section of the cluster settings.
In this design, the DRS Automation level is set to fully automated so when VMs are powered on they're automatically placed onto hosts with sufficient capacity. The VMs are also automatically migrated from one host to another to load balance the cluster. Additionally, the migration threshold of the DRS cluster is set at the midpoint between conservative and aggressive such that priority 1, priority 2, and priority 3 recommendations are applied. vCenter Server provides at least good improvements to the cluster’s load balance.
The following table shows the settings for the vSphere DRS cluster in the VMware vSphere Web Client.
Setting | Value |
---|---|
Turn on vSphere DRS | Selected |
Automation Level | Fully Automated |
Migration Threshold | Apply priority 1, priority 2, and priority 3 recommendations |
Enable individual machine automation levels | Selected, set to 15 ms |
For more information about configuring these settings in the vSphere Web Client, see the VMware vSphere documentation.
Along with the automation level and migration threshold of the cluster, this design enables VM automation so that you can override values per individual VM. More granular control of VMs enables further prioritization of the load balancing of VMs.
Power management
When the VMware Distributed Power Management feature is enabled, DRS compares cluster- and host-level capacity to the demands of the cluster’s VMs, including recent historical demand. The power management feature places or recommends placing hosts in standby power mode if sufficient excess capacity is found or powering on hosts if capacity is needed. Depending on the resulting host power state recommendations, VMs might need to be migrated to and from the hosts as well.
In this design, power management is not enabled as no operational or financial benefit exists to powering on and off hosts in the cluster.
vSphere HA
vSphere provides HA for VMs by pooling them and the hosts that they are on into a cluster. Hosts in the cluster are monitored and if a failure occurs, the VMs on a failed host are restarted on alternative hosts.
In this design, vSphere HA is enabled with host monitoring and VM monitoring on the cluster.
Host monitoring
Host monitoring allows hosts in the cluster to exchange network heartbeats and enables vSphere HA when it detects failures. This feature is enabled in this design.
Virtual machine monitoring
The VM monitoring feature uses the heartbeat information that VMware Tools captures as a proxy for guest operating system availability. VM monitoring allows vSphere HA automatically to reset or restart individual VMs that no longer have their ability to heartbeat. This design enables both VM and application monitoring.
Failure conditions and VM response
The failure conditions define how the VMs fail and the response that is given to each failure. In this design, the VM restart priority is set to medium. Review this value and adjust settings so that the restart priority matches the importance of the workload. Additionally, the response for host isolation is set to “Power off and restart VMs” so that VMs are not affected by an isolated host in the cluster. The rest of the values for this setting are set to default.
The following table shows the settings for the vSphere HA cluster in the VMware vSphere Web Client.
Setting | Value |
---|---|
VM restart priority | Medium |
Response for Host Isolation | Power off and restart VMs |
Response for Datastore with Permanent Device Loss (PDL) | Disabled |
Response for Datastore with All Paths Down (APD) | Disabled |
Response for APD recovery after APD timeout | Disabled |
VM monitoring sensitivity | Custom |
Failure interval | 50s |
Minimum uptime | 90s |
Maximum per-VM resets | 10 |
Maximum resets time window | Within 1h |
For more information about configuring these settings in the vSphere Web Client, see the VMware vSphere documentation.
Admission control
vCenter Server uses admission control to ensure that sufficient resources are available in a cluster to provide failover protection and to ensure that VM resource reservations are respected. In this design, the failover capacity is reserved by specifying a percentage of the cluster resources. The defined failover capacity is set to 25% CPU and 25% memory.
Datastore heart beating
vSphere HA uses datastore heart beating to identify hosts that failed and hosts that are on a network partition. Datastore heart beating allows vSphere HA to monitor hosts when a management network partition occurs and to continue to respond to failures that occur. In this design, the heartbeat datastore selection policy is set to “Automatically select datastores accessible from the host”.