Insufficient data from Andrew Fryer

The place where I page to when my brain is full up of stuff about the Microsoft platform

January, 2013

  • Evaluate This–VDI

    Microsoft is serious about Virtual Desktop Infrastructure, and the first sign of this in Windows server is when you try and add a role or feature..

     

    If you opt for install Remote Desktop Services Installation and select Virtual Desktops as in my short screencast you can see that a lot of work has gone in to making this as simple as possible.  However there is more to VDI in Windows Server 2012 than a good installation experience for example:

    • Hardware Optimisation with RemoteFX.  Making use of spare CPU capacity on your hosts to spoof a powerful graphics card into the desktops for multi-touch, and multiple screens
    • Dynamic memory has start up memory settings as an option so you can quickly start a Windows 7/8 VM with a lot of memory then peg it back to a lower running value until the VM is under memory pressure when it can get more subject to priorities you set.
    • User Profile Disks.  These are differencing disks where each users state and profile are stored.  These differences are based on the gold image you use to create your VDI environment for example a Windows 8 machine with Office 2010 on.  The clever bit is that if you wanted to revise or patch this gold image, say to put Office 2013 on, then you can do this without users’ losing their profiles and they’ll get the new version once they log off and log in again.

    Notes:

    In this screencast I put all of the  Pooled VDI virtual machines’ storage onto a highly available file server (this  post shows you how I built that) and this is where my user profile disks are also stored so that no matter which physical host a user gets their pooled desktop from they will still get their own user settings.

    I used a separate VM for each role in my remote desktop infrastructure, however if you elect for a quick setup then you can have all the roles on the one physical host from which the virtual desktops will run as well.

    There’s a more details lab guide here, and you can easily navigate to other labs form here for a quick setup as well.  Either way you’ll need an Evaluation Copy of Windows Server 2012 and Windows 8

  • Evaluate This – Highly Availability File Servers

    Storage in Windows Server 2012 is more than firing  up a few file shares and setting up security on them. With SMB3 file shares can now be used to host high performance application data like running hyper-V virtual machines and SQL Server databases.  Speed is one thing but reliability is what really matters and in the server world that means high availability and in Server 2012 you can now create a file server role in a cluster.  The nodes (up to 8 for this role) need to see some sort of shared storage but not necessarily a SAN. As you can see in this short screencast it’s a simple exercise and the share can be configured for a variety of uses including NFS.

    Notes:

    This role also uses the clustered shared volume technology introduced in Windows Server 2008 R2 for live migrations in Hyper-V. Actually you must use CSV if you are using the share for application data as this hands off the read access to another node in a cluster fast enough to enable continuous availability.

    In my demo I did everything in VMs except having an iscsi target on my host, and will this is great to show it working and to evaluate the technology. However this really gets interesting if you create something called a collapsed cluster. The hardware vendors are working on a cluster in a box where several computers in a commodity racked box are cross wired to SAS/JBOD disks.  You could then create a clustered file share on this and put your virtual machines into the same cluster and then they too would be highly available as there would be no single point of failure in this box, and you wouldn’t need a SAN to do it.  I’ll show you how this works in a separate video.

  • Evaluate This – Hyper-V Replica

    In many of the screencast in this series I have moved a VM around my demo setup, however there has only been the one copy of it whether it was on a scale-out file server, in a cluster or both. In any production environment you would want to augment this with additional disaster recovery techniques including have a backup of the key virtual machines somewhere.

    Replica in Windows Server 2012 is a partial answer to this.  You setup a  process to make an offline copy of a given virtual machine (VM) on another server and continually keep it updated. This replica VM can be updated over “UK speed” (don’t get me started!) broadband and you can also maintain up to 4 roiling snapshots enabling you to go back past a data error you may want to correct.   This screencast shows you how to set it up ..

     

     

    Notes:

    The replica is always off and it’s up to you under what conditions you invoke failover and of course you can script this in PowerShell with Start-VMFailover as well as all of the configuration for Replica I did in the screencast.

    The principal and replica can either be a cluster or an individual server. 

    In my demo all the servers belong to the same domain but if that’s not the case then you can use CredSSP to set this up. One use of this is that hosters are planning to offer replica as a service so you’ll be able to set your critical VMs to be replicated (is that English?) over the internet into their data centres as a service.

    As I briefly mention in the screencast you can also set your replicated VM to preserve it’s network settings when you fail over to it in its new location.

    You are going to need 2 x hosts/physical servers to try this and an Evaluation Copy of Windows Server 2012.  It doesn’t matter what OS your virtual machine is running, but do be aware of what applications etc. are supported for replication, e.g. SQL Server , System Center, SharePoint etc.

  • Evaluate This–Collapsed Cluster

    19 March 2013 -  This post has been changed to reflect the best way to configure a cluster in a box

    In my last post I used two clusters; one to host a high availability (HA) file server where I stored a virtual machine and another cluster to run the virtual machine.  The file server cluster was built from two virtual machines (VMs) and is commonly known as a guest cluster. However to enable HA for VM I needed to cluster two physical server (aka my Dell Precision Laptops). 

    What I could have done was to put both the File Server role and the VM role into the same cluster (which would still have to be built from physical servers),  However that configuration won’t work and isn’t supported and if you try it you’ll most likely run into Access Denied errors. For more on this look at this post by Jose Barreto, on the File Server engineering team.

    I mention all this because there is a new breed of hardware appliances coming to market known as Clusters in a box; So Simply take two Server motherboards each with CPU memory etc. stick them in a 2/4 AU box which also contains multiple power supplies network cards as well as a bunch of SAS disks again with multiple controllers so everything is redundant. A good example is this from Fujitsu..

    So how could use this new type appliance to create highly available  virtual machines (HAVMs)  based on Windows Server 2012 and Hyper-V?

    You would create cluster shared volumes over the built in storage  via storage spaces and so each node would have access to C:\ClusterStorage\Volume[x] and then create HAVMs on top of that.  I haven’t had the chance to create a screencast for that as I need a cluster in a box in order to do that (I’ll re-edit this post when I do) so in the meantime I would refer you to Jose’s other posts on Hyper-V over SMB3

    In the meantime if you are experimenting with Windows Server 2012 you can get the Evaluation edition here

     

     

     

     

     

     

     

     

    But you may be wondering why you would bother as this seems to be needlessly adding another layer of complexity and another potential source of problems rather than just using a SAN.  The answer is that a HA File Server doesn’t have to be built on top of a SAN it could be built on any disks you have including JBOD (Just A Bunch of Disks and  SAS (Shared Serially attached SCSI) disks. Hardware vendors are bringing out these cluster in a box appliances ; two servers SAS storage multiple controllers and network interfaces and a collapsed cluster like this is an ideal way to set these up to run lots of VMs in a small business that wants to run its own infrastructure.

    Notes

    The two roles (the VMs and the storage) can’t run on the same node (as per this post by Jose Barreto, a collegue of mine on the File Server team), but if you are doing maintenance on one node in a two node cluster then they will have to be.

    I used a small disk as a quorum disk which is needed to decide which node “owns” the cluster after a node fails the answer being the one that has ownership of the quorum disk.

    Nodes in a Windows Server Clusters need to members of the same domain.  Does this mean you have to have a physical domain controller outside the cluster in case of a cluster failure? No clusters in Windows Server 2012 will start without one but remember they  need to find each other and so you will need to use things like fixed IP addresses and an etc/hosts file in each node so this can happen before your DNS and DHCP infrastructure comes up.  You could also run a DC as a non HA VM on each node of the cluster and these only need modest resources (512Mb RAM 10GB disk etc.)

    While I used the evaluation edition of Windows Server 2012, I could have built all of this using the free Hyper-V Server 2012 and while you would still need to license any operating systems in the VMs with this, you can build collapsed clusters/cluster in a box solutions for production with this edition.

  • Evaluate This–Storage Migration

    In some smaller organisations virtual machines (VM) often run on local storage DAS – Direct Attached Storage on the hosts whereas in bigger businesses many if not all production VMs are hosted on shared storage (e.g. a SAN) , so the virtual machine executes on a given host but the virtual hard disk and VM metadata resides elsewhere.  In either case there might be times where you would want to move the storage for a virtual machine but leave it running on its current host.  For example you might want to move a virtual machine from DAS to a SAN as it becomes more critical to a business, or you are upgrading or replacing a SAN.  This is no a simple process in Windows Server 2012 and you can leave the machine running while you do it as you can see in this short screencast..

    where my poor SQL Server 2012 VM gets moved around my demo rig while running a complex query again and again.

    Things to note:

    This screencast moves the VM to a highly availability file share, which I created in an earlier post in this series.  Note that that file share is specifically designed to host running VMs using the new SMB 3 capabilities in Windows Server 2012, and configured to do so as opposed to storing conventional files or to run as an NFS file share.

    Permissions to that share are granted to the hosts running hyper-V in my case I created a group called Hyper-V Servers to put my hosts in and assigned permissions to that.

     

     

    You'll need a copy of evaluation copy of Windows Server 2012 as these advance features aren’t available from Hyper-V in Windows 8
  • Evaluate This–High Availability Virtual Machines

    Server virtualisation is all about decoupling the operating system form the hardware it’s running on, and one of the benefits of doing this is to ensure that a virtual machine (VM) can be made resilient to any underlying hardware failure.  In the world of Hyper-V this is achieved by building a Windows Server cluster and adding the VM as a role into that cluster.  From Windows Server 208 R2 this also gives the benefit of moving the virtual machine around  nodes on the cluster without stopping the virtual machine (known as live migration).

    In Windows Server 2012 you still need to use a cluster to make virtual machines highly available, but you also have the option to build a cluster without any shared storage using a file share to host the virtual machines storage and metadata.  This screencast shows how that works..

    Things to note.

    This builds on two other posts in this series:

    What I have done here illustrates the technology for high availability in Windows Server 2012  and is not a high availability solution itself – the high availability file server is running on two virtual machines but these are connecting to an iscsi target that isn’t highly available itself and I have no redundant network infrastructure.

    As with several of my screencasts it’s a SQL Server 2012 VM that's is being migrated around.  I run my Resource Governor demo application on the VM while it’s being migrated as this enables me to max out the CPU on the VM to show that migration doesn’t significantly slow this process and certainly doesn’t stop it.  I also use remote desktop to connect to the VM because if I used the VM console it would drop during migration because the console is connecting to the VM via the host and of course the host changes during the migration.

    To try this yourselves you’re going to need at least two physical hosts (laptops/servers etc.) as well as Windows Server 2012.

  • Evaluate This - Shared Nothing Live Migration

    In earlier versions of Windows Server you needed to build a cluster with shared storage (i.e. a SAN) if you wanted to move a virtual machine from server to server without stopping it (known as Live migration in Hyper-V).  In server 2012 you just need to configure Live Migration in each of the servers as per this screencast..

    But why does this matter?  in a word -  agility.    Particularly for smaller businesses who don't have the budget or expertise to run a SAN, and for whatever reason want to manage their services in house rather than use the cloud.  Key services can be moved around as needed without stopping them and this means that planned maintenance tasks can be carried out during the working day.

    Setting this up is really easy and we usually get our delegates at our IT Camps to pair up and do this using their own laptops without too many problems. If you have two desktops/laptops lying around you can get and an evaluation copy of Windows Server 2012, and follow along.

     

    Things to note.

    The number of live migrations you configure is up to but if you only have limited networking you’ll set this low as you don’t want to interfere with access to the VM if that traffic is on the same network (You can set live migrations to use specific IP addresses).

    You can use CredSSP or Kerberos (i.e. the host machines are in the same domain) to setup the trusts between the hosts for this to work. Note the domain etc. of the virtual machine isn’t relevant

    There is no high availability here – If the host running the VM stops working so does the VM and if the host suffers a disk crash the virtual machine will be gone as well, so this technique just helps with planned maintenance.