About 5 months ago, I made a video about configuring the network load balancing cluster in Windows Server 2008. I am continuing the series about clustering the Windows Server 2008 with the next type of clusters – the failover cluster. Also known as “high availability” cluster.
Although Windows Server 2008 supports 4 types of clusters – Network Load Balancing, Failover, Computational and Grid, the most commonly used are the first two. Also, we’ll talk about the private clouds later, as they are doing similar job, but in Windows Server 2008 the private cloud is the functionality of an application called System Center Virtual Machine Manager 2012, so it’s not the system core feature, such as clustering.
During the series of demos we’ll talk mainly about failover and network load balancing clusters, as the High Performance Computational cluster requires it’s own special edition of Windows Server 2008, called Windows Server 2008 R2 HPC Edition, and chances are – you won’t ever have the requirement to set up such environment.
Windows Server 2008 Cluster Categories
As you can see in the following slide, there are two categories of clusters by the way they share resources.
Failover cluster belongs to the second group, which means that it is a group of computers, where only one node (i.e. the machine participating in the cluster) owns the resource. You may have two or more machines working as nodes in your failover cluster, but only one of them will serve clients at any moment of time. Once that machine fails, another node takes ownership of resources (shared drive, for example) and starts serving clients instead of the failed node.
The main difference between the Network Load Balancing (NLB) and Failover clusters is that in NLB cluster all nodes are serving client requests, so it allows you to scale up your web server to what is called the web farm. NLB cluster can only serve sessionless applications, like HTTP or DNS queries.
The HPC cluster, on the other hand, makes all nodes work as parts of the whole, adding power to a virtual supercomputer. The grid cluster differs from HPC by that each node is doing it’s own job separately from other nodes.
Windows Server 2008 Failover Cluster Features and Requirements
Unlike in the NLB cluster, the failover cluster’s node “owns” the cluster, and another node could only take it from it’s cold dead hands.
Just like in NLB, you should have two networks – one for communication with clients and one for heartbeat traffic, which nodes use to exchange cluster-related information.
Like the NLB cluster, all nodes in failover cluster should have the same or compatible hardware and the same configuration. Microsoft insists that all components should have “Certified for Windows Server 2008” logo, but the key is that you won’t get support from Microsoft, if components are not marked as such. Logo by itself doesn’t make your hardware work better.
Make sure you update the firmware of your hardware components so they are the same on each node. This could cause the problem on successor node, which you’ll notice only when the node failure occurs, and our goal is to make node failure the only problem we have. The chance to catch such glitch is minimal, but still.
The key in failover clusters is to eliminate the single point of failure. This means that the same as your nodes are dubbing each other, it would be beneficial to have the network communication redundancy as well. In case of your network switch failure or any single cable malfunction, your solution must be able to switch without delays.
Of course, you only have to invest in such measures if server uptime is paramount. Having reserve network switches is not a common practice even for large IT companies. However, the point is that Windows Server 2008 is not the end in your quest for complete failover.
The Storage for Windows Server 2008 Cluster
If you have SAN (Storage Area Network, that is) – use it. If not – at least use the file server with RAID, because if you have failover cluster which is dependent from a file server which could easily fail – you don’t have a failover cluster.
SAN is expensive, but the most reliable solution for storage, so if your budget permits – just get it, at least one, for the most important data.
However, in video I am using the share folder on file server, just because most failover clusters are created by companies which have no reliable SAN systems in their network, so I don’t want to show things you are not likely going to use.
Besides, the next part most likely will be devoted to setting up iSCSI, which is the preparation phase for setting up the failover cluster. Then we’ll talk about management of failover clusters and look at what’s new in Windows Server 8.
Update 25/02/2013: there is a new article on failover clusters in SQL Server 2012, about AlwaysOn Availability Groups.
Configuring Failover Cluster in Windows Server 2008