Hi All 
...
In the thirteenth part of our series, we'll go through second main Clustering feature in vSphere: High Availability (HA) as well as, Admission Control and Slot Size Calculations which are considered the hardest parts in vSphere HA Clusters designing.
Credits:
- Duncan Epping
Now, Let's Start...
1. What’s New in vSphere 5.0 and later HA:
The following articles by Duncan Epping describes enhancements and changes in HA in vSphere 5.0, vSphere 5.1 and 5.5 respectively:
vSphere 5.0: http://www.yellow-bricks.com/vmware-high-availability-deepdiv/
vSphere 5.1: http://www.yellow-bricks.com/2012/09/12/whats-new-vsphere-5-1-high-availability/
vSphere 5.5: http://www.yellow-bricks.com/2013/09/04/vsphere-5-5-nuggets-high-availability-enhancement/
2. HA Admission Control and Slot Size Calculations:
Admission Control is a feature used by vCenter Server to ensure that sufficient resources are available in a cluster to provide failover protection and to ensure that virtual machine resource reservations are respected. This feature and its calculations done are based completely on a certain concept which is Slot and Slot Size. Slot Size is the equivalent amount of CPU & Memory for an average VM (Slot). According to the number of slots calculated, the cluster determine how much resources can be reserved for HA in case of host failure.
CPU of a slot: Largest CPU reservation in the cluster or its default value is 32 MHz if there’s no reservation.
Memory of a slot: Largest Memory reservation in the cluster in addition to overhead or its default value is 0 MB in addition to overhead.
Number of slots per host: Min. [{(Total CPU of the host root’s resource pool (listed on Resource Allocation tab when connecting directly using vSphere client)/CPU of a slot}, {Total Memory of the host root’s resource pool (listed on Resource Allocation tab when connecting directly using vSphere client)/Memory of a slot)}].
In vSphere 5.1 and later, changing default and max. slot sizes can be done through web console or via advanced HA Settings:
Setting | Description |
das.vmCpuMinMHz | Default CPU slot size in MHz. |
das.vmMemoryMinMB | Default Memory slot size in MB. |
| das.slotCpuInMHz | Max. CPU slot size (overrides max. reservation) in MHz. |
das.slotMemInMB | Max. Memory slot size (overrides max. reservation) in MB. |
Keep in mind that even if the aggregate resources across the cluster is enough to satisfy a VM requirements to be powered on after failing over, that VM can’t be powered on if there’re no enough resources on a single host to satisfy its requirements as these fragmented resources aren’t considered by the Admission Policy.
For more deep-dive information, check the following series of articles written by Duncan Epping:
http://www.yellow-bricks.com/2012/06/18/ha-admission-control-basics/
http://www.yellow-bricks.com/2012/06/20/ha-admission-control-basics-part-22/
http://www.yellow-bricks.com/2012/07/03/answering-some-admission-control-questions/
http://www.yellow-bricks.com/2009/08/12/ha-and-slot-sizes/
http://www.yellow-bricks.com/2009/10/06/slot-sizes/
3. Specifying HA Failover Hosts:
When selecting certain hosts to be failover hosts in a HA Cluster, DRS will not try to balance failover hosts and VM-VM affinity rules won’t be applied.
4. Cluster Settings - Datastore Heartbeat:
Datastore Heartbeat is a supplementary feature in HA Cluster that helps Cluster hosts detect which ones are down and which are up in case of network failure through certain heartbeat sequence written on the shared datastore seen by the cluster.
When the master host in a vSphere HA cluster can’t communicate with a slave host over the management network, the master host uses datastore heartbeating to determine whether the slave host has failed, is in a network partition, or is network isolated. If the slave host has stopped datastore heartbeating, it is considered to have failed and its virtual machines are restarted elsewhere.
This official article by VMware describes how Datastore Heartbeating feature is working:
Number of datastores recommended to use Datastore Heart-beating is two, but it’ll work efficiently with only one datastore although it’ll give an error which can be ignored using an advanced setting (das.ignoreInsufficientHbDatastore). In addition, the number can be controlled by an advanced setting (das.heartbeatDsPerHost).
By default, VMFS datastores will be selected rather than NFS and two datastores from different storage arrays will be selected.
5. Network Isolation Response:
Network Isolation happens when a host lost his management network and can’t connect to its cluster master host while it continues to run. At this point, certain action must be taken to guarantee successful operation of Cluster and its VMs and here comes Network Isolation Response. Available actions to be taken are: Leave powered on (default), Shutdown VM or Power-off VM.
The following article from VMware describes Network Isolation Response in details:
6. Cluster Settings – VM & Application Monitoring:
VM Monitoring is a feature in HA Cluster which monitors VM activity through VMware Tools and restart VM in case of halting (No heartbeat or any IO activity received).
Application Monitoring is another feature in HA Cluster that monitors certain applications inside the VM using also VMware Tools and certain driver installed in the application. It restarts the VM itself in case the application halts or doesn’t respond (No heartbeat received).
This official article by VMware describes how sensitivity of VM Monitoring and Application Monitoring features works:
7. HA Cluster Advanced Settings:
The following KB article by VMware lists all available advanced settings in HA Cluster:
Share the Knowledge ...
Previous: vSphere 5.x Notes & Tips - Part XII: