Microsoft Hyper-V Cluster Design
上QQ阅读APP看书,第一时间看更新

Configuring nodes

To begin configuring a node, it's helpful to have first documented how you expect the node to be configured. Most importantly, you'll need to have decided what the computer name will be and what IP addresses it will have. If you will be employing NIC teaming, that should be clearly documented. Connections to physical switch ports should be documented and the physical switch should be prepared. Unless you intend to place all physical NICs into a single team, you may choose to connect them later so that you can match the logical adapters that Windows sees to the physical adapters.

You can configure a node through the GUI or through command-line and PowerShell prompts. If desired, you can even mix GUI and shell methods. Unless your node is running a version of Windows Server with a GUI, you'll have to run these visual tools from a remote system. To do that, the node will need to be attached to the network and have its security properly adjusted. For non-GUI installations, it is highly recommended that you use PowerShell for initial setup.

Initial node configuration using GUI tools

When your freshly installed server boots, you will be presented with Server Manager. Begin on the Local Server tab. The initial steps are standard operations that would be performed on any Windows Server installation. The options of primary interest are indicated in the following screenshot:

Click on the blue text on each item to be taken to the window where the option can be changed. Be aware that renaming the system will require a reboot. Renaming can be delayed until the end of these steps and done at the same time as the domain join, if desired. The examples in this book use the naming convention as used by the fictional Techstra company from Chapter 1, Hyper-V Cluster Orientation, so further screenshots in this section will show it with a name of sv-hyperv1.

Using the GUI to configure networking

The natural sequence is to configure networking immediately after the preliminary setup. If you are planning to use virtual network adapters in the management operating system and you want those to be in the initial configuration (for example, if you are planning to use a virtual adapter for the management role), then skip forward to role installation and return after the Hyper-V role has been enabled. The virtual switch and virtual adapters are not available until that has been completed.

Renaming network adapters

First, identify the physical network adapters. On the Local Server tab in Server Manager (shown previously), click on the blue text after any of the network adapters to be taken to the network adapter configuration screen. Ensure you use consistent names for your adapters so that they are easily identifiable. Preferably, use a name that allows you to determine which physical adapter it represents. The simplest way to determine this is to either unplug the physical cable or disable its switch port in the switch's interface. Conversely, you can start with all adapters unplugged and connect them one at a time while you rename the adapters. The following screenshot shows an adapter being renamed:

You can rename an adapter by highlighting it and pressing F2 or by right-clicking and choosing Rename.

Note

It is highly likely that you will be typing these names in PowerShell commands, so it is recommended that you use a convention with short and memorable names.

In the R2 series of products renaming the adapters may be unnecessary due to a new feature called Consistent Device Naming (CDN). If Windows is able to determine names from compliant hardware, it will automatically use them. They will show up with designators such as NIC 1 and SLOT 5.

Creating network teams

Next, you'll want to create any teams that you intend to use. On the Local Server screen in Server Manager, click on Disabled next to NIC Teaming. You'll be presented with the Nic Teaming window. Click on the Tasks drop-down list in the Teams section and click on New Team. The following screenshot shows this:

In the screen that appears, first give the team a name. If you are teaming adapters for a specific purpose, such as Live Migration, name the team accordingly. If you are teaming for converged fabric (this concept is explained in detail in Chapter 5, Network Design), use a generic name. Next, check the adapters that will be used in this particular team. Finally, click on the drop-down list for Additional properties and set these items accordingly. If you're not certain what to choose, pick the Switch Independent teaming mode, as the other two require special settings on your switch. Clicking on the link after Primary team interface allows you to set the traffic on the new logical adapter so that it is tagged for a specific VLAN. Unless you have a specific need of this, it is recommended that you leave it in the default VLAN.

If you want the team to participate in a specific VLAN, it is preferred to use your switch's interface to place the switch ports in a native VLAN instead of tagging in the operating system. VLAN tagging at the team level is strongly discouraged if the team will host a Hyper-V virtual switch.

Once the team is created, it will appear in the previously empty boxes in the NIC Teaming window. You can double-click on these new objects to make any necessary changes. After creating or making changes to an LACP team, it is normal for it to show LACP negotiation errors for a short time after creation. If these don't disappear after a few minutes, check your physical switch configuration. Not all switches support all possible LACP methods. As with converged fabric, the teaming mode and load-balancing modes are explained in greater detail in Chapter 5, Network Design. They can be modified later, so don't worry too much about making an incorrect decision.

The creation of a team also makes a new logical adapter to represent it. This is now visible in Server Manager (after it refreshes) while the physical adapters that comprise it are now hidden. If you access the Network Connections screen, it will appear there as well, and you can configure it like any other adapter. Notice that the member physical adapters are still visible in Network Connections but are now unbound from everything except Microsoft Network Adapter Multiplexor Protocol. If the team will be used for a single purpose, you'll make any necessary changes (like IP addresses and DNS server entries) on the new logical adapter. If you'll be using the team to host a Hyper-V virtual switch included as a converged fabric, don't assign IP information.

Enable roles and features

After the above options have been set, configure the system's roles. In Server Manager, you can click on Manage and then Add Roles and Features or you can click on 2 Add roles and features from the dashboard. Screenshots are included previously in the section on adding the management tools to a Windows Server installation. This time, you'll want to select Hyper-V on the roles page and Failover Clustering on the features page. If you haven't already installed the related management tools, you will be prompted upon choosing the role or feature.

When installing Hyper-V as a role in this fashion, additional screens are added to the wizard. The first is simply informational. The next allows you to create a Hyper-V virtual switch. It is highly recommended that you create the switch later. If you create it now, you will not be able to set SR-IOV or the QoS method. The switch must be destroyed and recreated to set either. If you choose to create the switch now, you must ensure that you do not select a physical adapter that is part of a team. Instead, pick the logical adapter that represents the team as shown in the following screenshot:

The next screen is for non-cluster Live Migrations. Since you will be creating a cluster, you can simply skip past this screen. However, cluster members can still perform Shared Nothing Live Migrations to each other and to any other Hyper-V Server hosts in this or a domain with a proper trust configuration, so you may elect to set these options now. Of course, they can also be set later, so there is no wrong way to configure this screen.

The third and final new screen will involve storage of virtual machines. Ideally you'd like to give these targets a shared location. However, this node doesn't yet know about those locations. They can be changed later by accessing the Hyper-V Settings screen of the node in Hyper-V Manager. Proceed through the rest of the wizard.

When the installation completes, you'll need to reboot the server. You can do this by switching to the All Servers tab, right-clicking on the server, choosing Shut Down Local Server, setting the options accordingly, and clicking on OK.

Note

The server will automatically restart two times after enabling the Hyper-V role.

If you chose to create a switch during the Hyper-V role installation, the management operating system will now have a virtual adapter on that switch which has any of the IP settings that were set on the adapter you assigned for the switch. If you don't want this adapter, the later sections of this chapter on Hyper-V Manager and PowerShell will show you how to remove it.

Creating or modifying a virtual switch

Before you can create a virtual switch, you'll need to know the Description of the network adapter you'll create it on. Open Network Connections. Each adapter will have a description field which is light gray in color in the default view. It will contain the hardware label, and if there are multiple adapters of the same manufacturer and type, a number. Since this is a long name, you can move the mouse cursor over each adapter to show the tooltip with the full name. You can also switch the view to Details mode. Refer to the screenshot in the previous section, Renaming network adapters, for examples.

Open Hyper-V Manager. Highlight the node in the left pane. Right-click on it and choose Virtual Switch Manager. Alternately, click on the link with the same name in the right pane. New virtual switch should already be highlighted. Ensure that External Switch is highlighted in the right pane and click on Create Virtual Switch. The following screenshot shows the resulting dialog with the information filled in:

You need to provide a name for the switch, which is vSwitch in this scenario. Under External Network, you must select the Description of the adapter you want to create the switch on. In this case, the Microsoft Network Adapter Multiplexor Driver is the description of the teamed adapter that was created in the preceding steps.

If you click on the checkbox to Allow management operating system to share this network adapter, a new virtual adapter will be automatically created and assigned to the management operating system. All settings from the selected adapter will be transferred to this new virtual adapter. The selected adapter will be transformed into a switch and will no longer have any TCP/IP settings of its own. You do not need to select this option now; it can be selected at any time. You can also use PowerShell to add virtual adapters (shown in the PowerShell portion of this chapter). They will be discussed in Chapter 5, Network Design.

As the dialog warning text indicates, you must select SR-IOV now if you wish to use it. It cannot be changed later. This checkbox has no effect if your network hardware does not support SR-IOV or if you create the switch on a team. SR-IOV is given in detail in Chapter 6, Network Traffic Shaping and Performance Enhancements.

Creating virtual adapters for converged fabric

This part is optional, and you will only need to follow it if you intend to use converged fabric. Please review Chapter 2, Cluster Design and Planning, if you need an explanation of this technology, or Chapter 5, Network Design, if you need more in-depth information. In order to make converged fabric work, you must create virtual adapters that are assigned to the management operating system. Currently, there is no method within the GUI to create these adapters. Please refer to the PowerShell section given later in this chapter, create the necessary adapters, and continue with the next section.

Setting IP addresses for management operating system adapters

This is performed in the Network Connections window, and the process is the same as configuring static IP addresses on any other Windows installation. However, only the adapter you have designated for the management role should have assigned gateway and DNS addresses. Others should have only an IP address and a subnet mask. Furthermore, all adapters aside from the management adapter should be excluded from participating in DNS. More explanations will be provided in Chapter 5, Network Design.

If you don't recall how to get to the Network Connections screen, look under the Rename Network Adapters section. Right-click on an adapter and click on Properties. Double-click on the Internet Protocol (TCP/IP) with the version in use on your network (typically version four). Enter the IP address on the first page. To disable the registration of an adapter in DNS, click on the Advanced button, switch to the DNS tab, clear the checkbox titled Register this connection's addresses in DNS, and click on OK all the way back to the Network Connections window.

The management network and the Live Migration network must have their own IP addresses. It is recommended but no longer required that you have a network dedicated to the cluster communications role as well. These IP addresses must be in separate subnets. These IPs must also be on distinct physical or logical (team or virtual) adapters. Additionally, if your connection to shared storage is over an IP network, you will need at least one address and adapter for that, also in a separate subnet.

Joining the computer to the domain

Joining the management operating system to the domain is performed just like any other Windows Server installation. You can start on the Local Server page in Server Manager and click on the workgroup name to get started. If you didn't rename the system at the beginning of these steps, don't forget to do so.

Initial node configuration using PowerShell

All of the same steps used in the GUI configuration method can be performed at a PowerShell prompt, as can some that cannot be performed in the GUI. Not all possible options for each command will be shown here. The entire PowerShell library is available on Microsoft's TechNet site.

For Windows 8 and Windows Server 2012, use the following URL:

http://technet.microsoft.com/en-us/library/hh801904.aspx

For Windows 8.1 and Windows Server 2012 R2, use the following URL:

http://technet.microsoft.com/library/dn249523.aspx

Unless otherwise specified, all commands are entered using a PowerShell prompt running with administrative credentials. This is the default mode for Hyper-V Server and Windows Server in Core mode. To access PowerShell in either of these installations, just type powershell at any command prompt and press Enter. For a GUI installation of Windows Server, you must be running at an elevated prompt. To initiate that, right-click on the PowerShell icon and click on Run as administrator.

Tip

Type logoff at the command prompt to end your session. If you accidentally close all command prompt windows, press CTRL + ALT + DEL (CTRL + ALT + END in an RDP session) and click on Task Manager. Click on File, then New Task (Run). Type CMD in the Open field and press OK.

Remember that PowerShell is a complete shell, just like the existing command line. It has its own rules for handling input and parameters. So, even though you can call on the same commands from PowerShell that you can from a command line, they may not always behave the way that you expect them to. Familiar DOS commands, such as dir, may behave differently than expected as they are actually aliases for PowerShell cmdlets.

If you are completely new to PowerShell, you can still use these directions. Newcomers are encouraged to become acquainted with PowerShell first. You can begin on the related TechNet page: http://technet.microsoft.com/en-us/library/bb978526.aspx. To maximize clarity for those unfamiliar with PowerShell, the commands in this book will avoid positional parameters and shortened aliases with the exception of ft and fl (aliases for Format-Table and Format-List, respectively).

Tip

Any time you need to restart the computer from PowerShell, issue the Restart-Computer command. This executes instantly without confirmation.

Basic configuration

You can use the aforementioned timedate.cpl file at any prompt to change the system time and time zone. You can also use the Set-Date cmdlet to set the time, but there is no native PowerShell command to set the time zone. Your domain may issue a time zone and clock corrections through group policy, so you may elect to skip this portion. The command will look as follows:

tzutil /s "Central Standard Time"
Set-Date 1:40pm

At a regular command line or in PowerShell, you can enter tzutil /? for more information on using this built-in utility. The Set-Date cmdlet can be much more involved than is shown. Refer to the online help for more information.

You can rename the computer with the Rename-Computer PowerShell cmdlet:

Rename-Computer -ComputerName . –NewName sv-hyperv2

In the preceding command, the period after the ComputerName parameter indicates this computer. You will receive a message in yellow text indicating that you must reboot in order for the name change to take effect. You can choose to reboot now or wait until a later point. The computer name is unimportant until you join a domain.

Enable roles and features

The complete cmdlet for setting all necessary roles and features is shown in the following command:

Add-WindowsFeature –Name Hyper‑V, Failover-Clustering, RSAT‑Clustering, RSAT‑Hyper‑V‑Tools

If you are running Hyper-V Server, the Hyper-V role is already enabled, but there is no harm in naming a role or feature that is already installed. Adding the tools is recommended but optional since their functions can be performed remotely. You could use the IncludeManagementTools parameter to automatically install RSAT tools for indicated roles instead of specifying the tool names, although unlike the shown command, it will not install tools for roles that are already enabled.

Once the roles and features are enabled, a table will be displayed showing the outcome. If a reboot is required, this will be indicated in the Restart Needed column. If a reboot is necessary and you intend to use converged fabric, you will need to reboot before continuing. Otherwise, you may skip the creation of the virtual switch and return to that after completing the remaining configuration steps and rebooting the host.

Note

The server will reboot two times while enabling the Hyper-V role.

Using PowerShell to configure networking

Windows Server 2012 introduced many new PowerShell cmdlets for network adapter management. They replace most of the functionality of the deprecated NETSH.EXE tool.

Rename network adapters

Renaming network adapters in PowerShell is simple, and is shown as follows:

Rename-NetAdapter -Name "Ethernet 2" -NewName "Onboard 1"

You'll need to determine how Windows matches its adapters to the physical hardware. The easiest way to do this is to unplug all adapters except one or plug in all except one. The following command will show you which adapters are connected:

Get-NetAdapter | Where-Object "Status" –eq "Up" | fl Name

If you wish to see adapters that aren't plugged in, change Up to Disconnected. You can use the pipeline to automatically submit an adapter for renaming:

Get-NetAdapter | Where-Object "Status" ‑eq "Disconnected" | Rename‑NetAdapter ‑NewName "Onboard 2"

If you use the pipeline in a way that attempts to rename more than one adapter, only the first will actually get the new name. The rest will generate errors.

Tip

Network adapters can also be manipulated by using their interface indexes and descriptions using the InterfaceIndex and InterfaceDescription parameters.

Converged fabric

The next two sections will explain how to create the teams and switches that make up a converged fabric. If you prefer, there is a script that can automatically place all available physical adapters into a converged fabric for you. It is available from the TechNet gallery at http://gallery.technet.microsoft.com/scriptcenter/Create-ConvergedNetwork-2c98bc37. Converged fabric is explained in Chapter 5, Network Design, so read through that if you need more information.

Creating network teams

Creating teams in PowerShell is a quick and simple process. You need to know which adapters you want in the team, what mode you want the team to use (Switch Independent, Static, or LACP), and the load balancing algorithm (Hyper-V ports, transport ports, IP address hash, or MAC address hash—R2 introduces Dynamic). You'll notice that there are more load balancing options available through PowerShell than there are in the GUI. The different options are covered in Chapter 5, Network Design. To create a team, use the following command:

New-NetLbfoTeam –Name "ManagementTeam" ‑TeamMembers "Onboard 1", "Onboard 2" ‑TeamNicName "Management" ‑TeamingMode SwitchIndependent ‑LoadBalancingAlgorithm TransportPorts

Remember that when you are entering parameters that have predefined options, such as TeamingMode and LoadBalancingAlgorithm, you can use TAB to cycle through their options.

The name you provide for the TeamNicName parameter will become the name for the logical adapter that represents the team. Choose a name that is short enough that it won't be burdensome to type into further PowerShell commands but which clearly indicates what function the team serves. If you do not provide this parameter, the logical adapter name will be identical to the team name.

This sample creates a team on two adapters, presumably for management, and leaves other physical adapters available for other purposes such as other teams. The examples throughout this book combine available adapters into a single team called vTeam.

Create a virtual switch

Unlike the GUI role wizard, adding the Hyper-V role by PowerShell does not automatically create a virtual switch.

Note

System Center Virtual Machine Manager sees these as standard switches. You must use SCVMM to create logical switches with complete SCVMM functionality.

The following command will create an external virtual switch that can be used for both converging management operating system roles and hosting virtual adapters for virtual machines:

New-VMSwitch –Name "vSwitch" ‑AllowManagementOS $false ‑NetAdapterName "vTeam" –EnableIov $false ‑MinimumBandwidthMode Weight
Tip

The SR-IOV and MinimumBandwidthMode parameters are completely optional, but they cannot be changed after the switch is created. All other options can be changed later. Chapter 5, Network Design, and Chapter 6, Network Traffic Shaping and Performance Enhancements, cover these settings.

The shown command creates a virtual switch and assigns it to an adapter with the name vTeam. Assigning a virtual switch to an adapter automatically makes it an external switch. You can create one external switch per adapter. That adapter can be physical or a logical adapter that represents a team. You cannot create a switch on a virtual adapter. When an adapter is assigned to a virtual switch, it can serve no other purpose. Switch types and details will be explored in Chapter 5, Network Design.

Setting the AllowManagementOS parameter to false prevents the command from automatically creating a virtual adapter. It does not block them from being added manually, as will be shown in the next section. If you omit this parameter or set it to true, a virtual adapter will be created that has the same name as the virtual switch. The IP settings that were assigned to the physical or logical adapter will be transferred to this new virtual adapter.

Tip

All cluster nodes must use the same names for virtual switches to ensure that virtual machines attached to them can move between nodes.

Create virtual adapters

The only way to create virtual adapters for the management operating system is to use PowerShell. The command structure is very simple:

Add-VMNetworkAdapter –ManagementOS –SwitchName "vSwitch" –Name "Management"

The ManagementOS parameter is a toggle; if not included, the system assumes you are trying to operate on a virtual adapter assigned to a virtual machine and will prompt you to provide the VM's name. This toggle appears in all virtual adapter PowerShell commands.

Virtual adapters in the management operating system live a dual life. For PowerShell commands and operations that are related specifically to virtual adapters, they are referred to by the name that you provided in the creation command. For general network adapter commands and applications, they are called vEthernet (the name used in the command). So, the preceding command would create an adapter named vEthernet (Management).

Tip

Only use Rename-VMNetAdapter to rename virtual adapters. It overrides any changes made using Rename-NetAdapter or from the GUI.

Assigning virtual adapters to VLANs

With the requirement that cluster nodes use separate subnets for their various roles, you may choose to place them in separate VLANs as well. If you are employing virtual adapters on converged fabric for these roles, the virtual switch will need to handle the VLAN tagging for those adapters. Set VLANs on your virtual adapters with the following command:

Set-VMNetworkAdapterVlan –ManagementOS ‑VMNetworkAdapterName "Management" ‑Access ‑VlanId 10
Setting IP addresses for management operating system adapters

The PowerShell command for IP address assignment is very straightforward. Only assign a gateway to the management adapter. Except for geographically distributed clusters, adapters for other cluster roles do not need, and do not benefit from, the ability to route. For those adapters, use the following:

New-NetIPAddress –InterfaceAlias "ClustComm" ‑IPAddress 192.168.30.10 ‑PrefixLength 24

The PrefixLength parameter refers to the subnet mask. This is given in CIDR notation; if you are unfamiliar with CIDR, consult with a networking expert or resource.

For the management adapter, which does need to be able to route, there is only one additional parameter, shown in the following code:

New-NetIPAddress –InterfaceAlias "vEthernet (Management)" ‑IPAddress 192.168.10.10 ‑PrefixLength 24 ‑DefaultGateway 192.168.10.1

There are a number of PowerShell commands that include the noun NetRoute for manipulation of gateway addresses. This topic is beyond the scope of this book.

In order to resolve the names of other computers and to register its own name, your system will need to know the addresses of DNS servers in your domain. Assign these addresses only to your management adapter:

Set-DNSClientServerAddress ‑InterfaceAlias "vEthernet (Management)" ‑ServerAddresses 192.168.1.20, 192.168.1.21

You do not want the other adapters to register their addresses with a DNS server as this can cause problems for other systems that try to communicate with the management operating system. Run the following command on those adapters:

Set-DnsClient ‑InterfaceAlias "vEthernet (LiveMigration)" ‑RegisterThisConnectionsAddress $false
Join the computer to the domain

The final major configuration change to be made during initial setup is to join the management operating system to your domain. This is a process that requires two PowerShell commands. The first sets up a PowerShell variable that holds the credentials that are used to join the domain. It is possible to completely script this. In this example, you will use a command that generates a pop-up box where you can type in the account information to join the domain. Remember to enter your credentials in DOMAIN\Account or account@domain format:

$JoinCredential = Get-Credential
Add-Computer –ComputerName . ‑Credential $JoinCredential ‑DomainName "Techstra.local"

Optional node configuration steps

You have completed the required steps for initial node configuration. After the system is joined to the domain and you have rebooted, perform any other steps that are particular to your standard process for domain member servers. Run Windows Update (use option 6 in sconfig.cmd on non-GUI installations). Install any monitoring agents or other such software. If you will be using iSCSI or Fibre to connect to storage, perform the preliminary connection steps as indicated by the storage manufacturer's instructions.

You may also choose to configure remote access settings on your cluster nodes. The most useful is Remote Desktop, but it is also possible to open up access for other tools as well. Because your nodes are part of Active Directory and you will presumably want to apply all settings equally, it is highly recommended that you use Group Policy to control these settings instead of manually applying them on each node. If you wish to enable Remote Desktop manually, it is enabled in the GUI as in all versions of Windows: on the Remote tab in Computer properties. It can also be set from option 7 in sconfig.cmd (other remote access can be enabled using option 4). You can also enable Remote Desktop from the command prompt with the following command:

cscript %windir%\System32\Scregedit.wsf /ar 0