I have been working with Provisioning Services (PVS) for a while and one of the questions that keep coming up is around Provisioning Services High Availability (HA).
What is HA? Maintaining access to a service or resource is probably the best explanation for this type of functionality.
What does HA provide? Within the context of the Provisioning Services resource, we are referring to the target devices ability to maintain an active connection to a vDisk. This is different to redundancy, which I will cover in another topic. This ability manifests itself in the customers’ expectation for the target device to maintain constant uptime. What we have seen in the latest release is instant availability (first error move to next PVS server providing vDisk) as opposed to failover if the system doesn’t recover itself.
Types of HA: There are many ways of doing it, they all work, but which one is right for you? That depends. Let’s take a look at the location of your vDisks.
Placing your vDisks Local on the Provisioning Server – “Distributed Store”:
Using the local hard disk subsystem of the Provisioning Server to store the vDisks provides the easiest way of implementing vDisk High Availability without additional cost.
What should we consider with this solution?
- Although it is easy to implement and maintain vDisks need to be manually synchronized between the PVS Servers.
- I/O performance depends on the capabilities of the hard drive subsystem (usually equal to NAS).
What are the recommendations? Network Interface Card (NIC) – Teaming should be used to increase the reliability and to I/O between PVS Servers and Target Devices. You can also add more PVS streaming servers and heavier reliance on load-balancing.
You might be thinking about the vDisk lock files. Would failover work properly redirecting the target device to another server without a vDisk access failure? Yes, vDisk locks should not be a problem. Let’s say you are using the vDisk in Private Image mode, the lock file holds the information of the Target Device accessing it. If it fails from Server 1 to Server 2, Server 2 should not have an associated lock file yet because no other vDisk has accessed it. It will then create a new one. If it fails back over to Server 1 from Server 2, the lock file will have the Target Device information and re-connect it. If you are using the vDisk in Standard Image mode the lock files should not prevent any read only access.
Now you might be concerned about the write cache. Server Side caching is not supported, unless the cache is re-directed to a share. If cache is being stored on the Server’s local drive and becomes unavailable after failover, the Target Device will eventually blue screen when it tries to access needed information on the cache file.
Follow me on Twitter: http://twitter.com/lizteixeira