VMware High Availability Sponsored by VMware

Speaker: Mike Laverick

Mike Laverick: Hi there! I’m Mike Laverick, and in this brief demo and video, I’m going to explain to you what VMware High Availability is, what its requirements are, and how you set it up. So, what is VMware HA? Well, it’s a single click to protect all of your virtual machines. It’s very, very easy to set up. If an ESX host fails because there’s been a serious hardware failure, what VMware HA will do is detect that ESX host has gone down and then restart your virtual machines on another ESX host, or a series of ESX hosts within the same cluster. It also includes a virtual machine monitoring feature as well. So, VMware can inspect the state of your virtual machine and the guest operating system inside that virtual machine, and even restart the virtual machine if it has failed. Optionally, on top of enabling HA, if you have the required licenses, you can enable features like the Distributed Resource Scheduler. This is I/O’s VMware’s vMotion feature, to move virtual machines around from one ESX host to another ESX host. On top of that, you can also enable the Enhanced vMotion Compatibility feature, or EVC. Now, this allows you to choose the latest features of Intel processors, for example, which have Intel VT and Intel Flex Migrate options on the actual chipset itself. What allows is for a mix of different processors, over a series of different generations, to coexist within the same DRS cluster. And, then, on top of that, you could even look at enabling VMware Fault Tolerance. Again, this is an attribute of the modern processors, especially the ones that have come from Intel recently, which allow something called a vBlock stack actually to exist. Now, this allows VMware to capture all of the activities that take place in one virtual machine and replay it back in real-time into another virtual machine, on another ESX host. So, the sky’s the limit really, in terms of offering even higher availability to the virtual machine than you’ll have with just VMware HA. The important thing to say is that HA is the building block of all of these technologies. So, if you get HA up and moving, then it’s a breeze to get these other technologies working as well. Before we actually do a demo and show you how easy it is to set up a HS, I’m actually doing a test of it as well. I want to talk about, briefly, what the requirements for HA to function, because a lot of these will be already in place when I do my demo. So, the first thing you need is shared storage. Now, what we mean by that is the virtual machine’s files, the VMX file, the VMDK file, that make up a virtual machine, must be on external storage. So, that could be either a fiber channel SAN, an iSCSI SAN, or an NFS file server. The files in the virtual machine must be accessible to the ESX host. It makes common sense, if you think about it. If an ESX host is running a virtual machine over here and it fails, then the files on the virtual machine have got to be accessible to the other ESX hosts that are still functioning for the failover to take place. The other thing that you’ll need is network redundancy. The way the failure is detected, if it actually happens, is by the ESX hosts all communicating to each other through that management network. That means if a network cable failed and there was no network redundancy to the ESX host, then you could get false positives, failover happening when you didn’t really want it to. In a way, we already do this in our environments. We plug two network cards into different network segments. We team network cards together so we’re protected from the failure of an individual network card. So, you may be doing

this already with other technologies, but it’s an important requirement to remember, as part of the overall setup. If you fail to set up network redundancy for HA, it will allow you to set up the actual cluster, but you will get a warning in every ESX host, indicating that step is being missed out. The other thing you’ll need to look for is similarly-sized hosts. Think about it for a second. If you have one server that has 8 GB of RAM and another server that’s got 32 GB of RAM, and the 32 GB of RAM server actually failed, how would you find all of the memory on another ESX host that has much less memory than the other ESX hosts? OK? So, a little bit of consistency, something that we all strive for in IT, in terms of the sizing of the ESX host, is common sense. To some degree, HA introduces a performance-capacity argument. If I lose an ESX host, and I’ve lost the server-mounted CPU and memory, there must be free and available CPU memory for the restart of the virtual machines to be both successful and also offer a decent quality of service to my end users. Finally, a common network configuration, so the virtual machine itself, when it fails over to the other ESX hosts, does not leave its network location. So, if you have any inconsistencies in your building of the ESX host, that might stop HA from working. Again, it’s another one of those consistency arguments. So, each of the ESX hosts, the closer they are, in terms of their configuration, in terms of them all being identical, the easier life will be, A, for management, but B, in getting acceptable results out of HA. The key to making HA sing and dance for your environment is to meet these requirements. If you fail to meet the requirements, then, as you may expect, the technology may not always deliver the results you’re looking for. Welcome back! You found me in my lab environment here. I’ve got two ESX hosts, one called ESX3, one called ESX4. I recently upgraded my lab environment to the newer Lenovo ThinkServer TS200 servers, and they need to gain access to more memory in my lab environment, but also because I needed the CPUs, these Intel Xeon X34030 CPUs will be supportive for advanced features, like VMware FT. Now, what I’ll be doing with these two ESX hosts is putting them in the HA cluster and then I’ll use the management card of the Lenovo servers to actually create a fictitious failure. I’ll be able to come into this management window and physically crash all of the ESX hosts, in this case ESX4, so you can actually see what happens. And, we see, in real-time, HA actually kick in and function. So, the first step is to create the cluster. That’s quite straightforward. You right-click the data center you’re working with and choose New Cluster. Give the cluster a name, such as Intel Cluster. I want to, at this stage, actually enable HA. That produces a lot of questions and answers that are required. I just want to get this cluster set up very quickly and show you the relationship between the virtual machines, the ESX host, and the cluster object itself. As a side note, not this time, I could, if I wanted to, enable EVC for my Intel hosts here. If you remember from the PowerPoint, EVC stands for Enhanced vMotion Compatibility. And, notice how the very latest generation of Intel Xeon processors are on the supported list. Now, in my case, with HA, I don’t really need EVC

enabled, but I just wanted to show you how the wizard will guide you through the process, if you want to enable additional features on your cluster. It is also possible to relocate the virtual machines files. That isn’t really relevant to us in this HA discussion, but you do have a global setting to do this relocation of what file is associated with the virtual machine. So, if I click Next and Finish, that should create the cluster. And, then it’s just a simple process of dragging and dropping the ESX host into the cluster, as I’ve just shown you there. So, if you wanted, the relationships of Virtual Center, it includes data centers that I could domain in Active Directory, if you’d like, and then you could have many clusters within that data center, across…it depends on how large your environment actually is. Next stage is just to enable HA for that cluster. And, so, as I said, a simple tick in the box, yes, there are many other questions that you could be asked at this stage, but I just want to enable HA. And, that should trigger the process of HA being enabled on both ESX hosts. It does take a little time for the HA agent to be installed and to be started, and it will take a little time because it must have the first ESX host as it creates the cluster, and then the second one joins the cluster. Here, in this time, you may notice that you get a little yellow exclamation mark appearing on the ESX host. That’s completely normal. It’s not an error. They’re still in the process of being configured and enabled by HA. And, so it does take a little bit of time for this process to complete. Well, the cluster has been created. It did take a little time for ESX3 to join the cluster, but this is the state that the cluster should normally be in, without any yellow exclamation marks or red exclamation marks. So, here’s where the fun stuff comes in. This is ESX4, and it has two virtual machines, mob01 and web01. And, it’s this particular ESX host that I’m going to crash. And, the ESX3 host is obviously going to be the host that receives those virtual machines, initially get registered and powered up on the ESX host. So, now is the moment of truth. I’m going to bring up back my tool that allows me to control the power state of the ESX host and I’m going to do a hard power off of that ESX host. So, that ESX host is no longer available. It will take a little bit of time for ESX4 to appear as being disconnected, not available to the cluster. There’s a couple of seconds while Virtual Center tries to communicate to the ESX host and say, “Are you still there? Are you still there?” But, within a relatively short period of time, it should come up as not being available. And, hopefully, if I leave the video running for long enough…there we are. See, the red exclamation mark has just appeared on that ESX host as being a possible problem? But, if I leave the window open, that host should eventually become disconnected and we should start to see virtual machines appearing on ESX3. There we are. We can see that that ESX host is now becoming disconnected. All I’m waiting for now is for the virtual machines that were on ESX4 to start appearing on ESX3. There we are. Now, those virtual machines have been restarted. This video has been trimmed slightly, to take out some of the pauses as I waited for these events to happen. So, the duration isn’t exactly as you would see in the real world. The other thing I would say is that the BCA client itself doesn’t automatically update every nanosecond. So, sometimes I use other tools to do this kind of monitoring. For example, I could have used a command line tool, which shows me, in a real-time view, what virtual machines are running on the ESX host. But, as you can see, HA has worked. And, as long as we’ve

met the requirements, it was a very simple process of creating a cluster, dragging and dropping a virtual machine to that cluster, and then making sure those ESX hosts restart the virtual machines properly. Well, that ends that brief demo about VMware HA. I hope you’ve realized how quick, how easy, how simple it is to created an HA cluster, when compared to other equivalent tools. I hope you got a real insight into the fact that simply because you are consolidating lots of virtual machines onto one ESX host, that you don’t have to worry that much about the physical failures anymore. We’ve got tools built into vSphere that allows virtual machines to be restarted whenever an ESX host fails, if ever an ESX host fails. Thank you very much for watching!