With the introduction of vSphere 5.5 there are two mayor HA improvements announced:
– vSphere App HA, on the Intarweb also known as App aware HA; High Availability at the application layer.
– vSphere HA detecting VM antiaffinity rules.
I’ll start with the latter.
HA detecting VM Anti-affinty rules
With vSphere DRS… Hey wait isn’t the subject supposed to be HA… Well yes, but the anti- or affinity rules are DRS rules. So a bit of DRS rule explanation;..these rules helps maintain the order of placements of VM’s on hosts throughout the cluster. Affinity rules are rules that places VM’s together on certain hosts. Anti-affinity rules are rules that places VM’s separate from those VM’s in the rule. Think of VM’s that are already in a software availability service, such as the nodes of a cluster. You don’t typically want the nodes on one physical host.
With vSphere 5.1 and earlier vSphere HA did not detect a violation of these rules (these rules are unknown to vSphere HA). After a HA failover the VM’s could be place on the same host, after vSphere DRS would kick in and vMotion the VM’s so the anti-affinity rules are satisfied (DRS needs to be in full automated to enable the auto vMotion). Applications with high sensitivity to latency would not like this vMotion and there is a (very slight) moment that HA application clustering service are at higher risk as both VM’s are on the same physical host. A failure of the physical host before the vMotion is completed, would impact a downed service.
In a application cluster service you could also choose to use VM Overrides to disable HA restart for the VM cluster nodes as the application service handles the application HA actions. After a failure you would have to manually get the failed node online (or add a new one) in the application service. But that looses automation…
With vSphere 5.5 HA has been enhanced to conform with the anti-affinity rules. In a case of a host failure the VM’s are brought up accordant to the anti-affinity rules without the need of a vMotion action.This enhancement is configured as an advanced option.
vSphere App HA aka App aware HA
We already have host and VM monitoring, with vSphere 5.5 lifts this to application monitoring. vSphere App HA can be configured to restart an application service when an issue is detected with this service. It is possible to monitor applications as IIS, MSSQL, Apache Tomcat and vCenter. When the application service restart fails App HA can also reset the virtual machine. Service actions can be configured with the use of App HA policies. VM monitoring must be enabled to use application monitoring.
App HA Policies are definitions of the number of times vSphere App HA will attempt to restart a service, the number of minutes it will wait for the service to start, and the options to reset the virtual machine if the service fails to start and to reset the virtual machine when the service is unresponsive. They can also be configured to use other triggers, such as e-mail notifications or vCenter alerts.
When a configured App HA policy is assigned to a specific application service, vSphere App HA is enabled for that service.
But what’s needed:
For App HA to work two appliances are needed in the environment (per vCenter), vSphere App HA and vFabric Hyperic. The latter is used by the App HA architecture to monitor applications and is a vFabric Hyperic Server that communicates with vFabric Hyperic agents.
The roles of the both appliances are as follow: the vSphere App HA virtual appliance stores and manages vSphere App HA policies. The vFabric Hyperic appliance monitors applications and enforces the assigned vSphere App HA policies. For monitoring the applications of a VM, vFabric Hyperic agents must be installed inside the VM’s of these applications. These agents are communication brokers for the applications of the VM’s and the vFabric Hyperic appliance.
The vFabric Hyperic agents are supported to be deployed at Linux and Windows os’ses for 32-bit or 64-bit applications. How and what is supported for vSphere 5.5 HA is not yet completely clear (service support for IIS6/7/8, MSSQL 2005/2008/2012, Apache Tomcat, Apache HTTP and vCenter). Following the current vFabric Suite supported OS’ses these include Windows 2003, Windows 2008R2, Red Hat Enterprise Server and Suse Enterprise Linux.
Well. Good Question. App HA is part of the vSphere Enterprise plus edition only. Costs of vSphere 5.5 is expected to be around the current vSphere 5.1 costs. But with what options, constrains and limits…..unknown. The General Availability of vSphere 5.5 is yet unknown.
Separately VMware vFabric Suite is currently available as a one-time perpetual license under which support and subscription (SnS) contracts can be renewed annually – See more at: http://www.vmware.com/products/vfabric/buy.html#sthash.lTBxCHHK.dpuf
How the both are combined at what options/editions/prices keep a look out for further vSphere 5.5 product announcements.
– Exiting. I have the HA BCO5047 – vSphere High Availability – What’s New and Best Practices in my Barcelona schedule to get some more insight at VMworld EU 2013.