In this video, Russ Long talks about some basic best practices for creating a cluster. Discover how to create solid cluster prior to enabling cluster options such as HA or DRS.
- [Teacher] Now before we dive into the details of an HA cluster, I want to make sure that we understand that the cluster itself is really important. We have to make sure that it is built in the best manner possible, otherwise we're creating problems before we even bring in the HA technology. So before you ever press the HA button, that HA option to turn high availability on, we have to make sure your cluster is created correctly. Now the linchpin to any cluster is Shared Storage.
This is the backbone where all the data is stored. Data will be moved back and forth several times to the Shared Storage option, so we have to make sure that our performance is up to snuff. Now there are many options for storage; iSCSI, Fibre, Fibre Over Etnernet, or NFS. Our choice here will directly affect the cluster at large. So make sure that your performance, your reads and writes, your IOPS, are all perfect before you turn on the HA option.
Now another key to building a good cluster is, I want you to make sure that your hosts within a cluster are built as similarly as possible. You want the same processors. You want the same type of memory. You want the same amount of memory to be in each host. This allows virtual machines to move from one host to another within a cluster without experiencing a big difference in the type of resources that are being offered. So a key to building a good cluster is to make sure the hosts are as similar as possible.
Now one of those main concerns that we should have when building a cluster is to make sure that we have the same vendor. You don't want to have AMD on one side and Intel on the other side of your cluster, it just won't work. Your migration is going to fail, and that's a bad thing. We have to have Intel-to-Intel migrations; otherwise, we'll have no migration at all. Now a lot of people said, "Well, EVC. "Enhanced V-motion compatibility. "That solves that problem, doesn't it, Russ?" Absolutely not.
EVC is for different models within the same family of processors and not different vendors. So it's for AMD-to-AMD processors or Intel-to-Intel processors that just happen to be different models. Even though we have EVC enabled, I would rather you use the same models within the same vendor. That is the best practice. EVC is when you have to shoehorn something in or just try and make it work with what you have. Always be mindful of the processors within your cluster and how you are going to utilize that cluster and how the virtual machines are going to move around inside.
Now last but not least on our cluster practices is to avoid placing virtual machines that have very high utilization on the same hosts within a cluster. This causes a contention for resources, such as our CPU, our memory, and even our networking. So we want to make sure that high-utilization virtual machines are kept separate on different hosts, even though they reside in the same cluster. A great example of this is our databases. If I have two servers that are databases within the same cluster, I'm going to try my hardest to make sure that those are separated at all cost.
This means when you place virtual machines within a cluster, you're going to plan out where those virtual machines reside on each host, rather than letting the chips fall where they may.
Note: This course will also help you prepare for the Configure and Administer vSphere Availability Solutions domain of the VMware Certified Professional – Data Center Virtualization exam. View the exam blueprint at https://mylearn.vmware.com/mgrReg/plan.cfm?plan=64180&ui=www_cert.
- How vSphere High Availability works
- The basics of clusters
- Understanding failure types and failure response
- Monitoring HA virtual machines and appliances
- Using heartbeats
- Creating and configuring clusters
- Configuring admission control
- Best practices: Networking, interoperability, and cluster monitoring