Note: this is a blog coming from a co-worker of mine – Tom Stephens – who is our HA guy – just like I am the SRM guy. He talks of an interesting product that works with HA to produce some very useful results. I hope you enjoy it!
A significant amount of interest has developed around Symantec’s ApplicationHA product. Of course, this interest has also generated a lot of questions. To help address these, the following will provide you with a brief introduction of what is the ApplicationHA solution, who can benefit from it, and how it works.
As customers started to move towards virtualizing their tier 1 applications, they had to address requirements that were not as significant when virtualizing applications in lower tiers. For these customers, ensuring availability of these tier 1 applications meant more than just ensuring the availability of the virtual machine. Just because the virtual machine was running, it did not mean that the application had been started. Likewise, just because an application's processes were present that did not mean the application was running properly.
Customers tried various solutions to solve this problem, with limited success. One solution tried was to install a clustering solution within the guest OS. This has several drawbacks, including:
An inability to use advanced virtualization features like DRS and vMotion, eliminating the ability to handle planned outages with zero downtime.
As these solutions do not have any insight into the virtualization infrastructure provided by VMware, they were unable to react to hardware or network issues properly.
The solutions were overly complex, requiring additional management of virtual machines to be used as fail over hosts.
The one-to-one relationship between an application and its virtual machine is broken. This means the virtual machine administrator needed to consult a separate console to determine where an application is running prior to taking maintenance actions on any given host.
Customized scripting solutions were also tried as a solution to this problem. The issues with this approach were that maintaining and updating the scripts induced unwanted management overhead and the scripts tended to increase all root cause analysis (RCA) activities.
Failing to find an adequate solution, these customers came to VMware. What these customers told VMware they needed was the ability to have a simple, easy to use solution that would provide them with visibility and management of their tier 1 applications while being able to enjoy all the features of virtualization, like VMware HA and DRS.
VMware took this feedback to heart and started looking for methods to solve this problem. After analysis of the problem, VMware reached out to its partners for assistance. Symantec was one of these partners, and ApplicationHA was born as a result.
VMware HA already provides a robust mechanism to detect failures of infrastructure components. This includes failures of the physical servers, the virtual machines, the state of the operating system, and so on. VMware HA does not, however, monitor the health of an application running within a virtual machine.
Symantec, on the other hand, has been a market leader in application clustering on physical servers for over a decade with products such as Veritas Cluster Server (VCS). They are experienced in monitoring an application’s state and reacting accordingly in the event of an application failure. Like other clustering products designed to operate in a physical environment, their ability to react to failures of infrastructure resources within a virtualized environment is limited.
The marriage of the capabilities from VMware and Symantec bridges this gap, allowing for a single solution that can react to failures of the entire stack, from the infrastructure to the application. At the same time, customers can continue to leverage all the benefits of virtualization, such as VMware HA and DRS.
From a high level view, the ApplicationHA product is comprised of two components. These include:
ApplicationHA Guest Components
The ApplicationHA Guest Components are installed within each virtual machine containing an application to be managed. In essence, this is simply a modified version of Symantec’s Veritas Cluster Server (VCS) product that runs as a single node cluster. It provides the ability to start, stop, and monitor an application and components (such as storage mount points, IP addresses, and so on) from within the virtual machine.
ApplicationHA Console
This provides an interface between the ApplicationHA Guest Components and vCenter Server though a vCenter plugin. This provides the ability to manage the ApplicationHA solution from within the vCenter management environment.
Once installed and configured, the ApplicationHA vCenter plugin provides a management view that shows the state of an application and it’s associated resources. From this view, administrators are able to perform a variety of actions, including starting and stopping the configured application and associated components. It is important to note that when ApplicationHA starts or stops an application it does so in a manner that honors any resource requirements that the application may have. For example, a Microsoft IIS server instance would require a storage location. In bringing the Microsoft IIS instance online, the storage needs to be available beforehand. ApplicationHA will ensure that each resource is brought online or offline in the correct order.
During normal operations, the application and the resources which it depends upon (file systems, IP addresses, and the like) are monitored every 60 seconds. A failure of the application or a resource it depends on is identified by the ApplicationHA Guest Components.
Depending on the settings configured, ApplicationHA can then restart the application and its associated resources a configurable number of times. If the a failure cannot be resolved by application restart, ApplicationHA can trigger VMware HA to restart the virtual machine.
Configuring ApplicationHA increases reliability of application components, while allowing VMware features such as VMware HA, DRS, and DPM to continue to function as expected. This provides efficient high availability services in the event of a failure of an infrastructure component.
The obvious benefits of this solution include:
The ability to react to infrastructure and application failures
No loss of other desirable functionality, such as DRS or DPM
A single pane of glass that can be utilized to visualize the health of the virtual environment and the applications residing within
Ability for administrators to simply manage a application, bringing it online and offline properly without assistance.
