Don’t disable the LPT port in the BIOS of VMware virtual machines if you used XenDesktop MCS with PVD.
Like you I have been really excited about PVD in XenDesktop 5.6. I think that controlling the space in the PVD is the magic. But with new deployments, comes new problems. Over the past few weeks I have been working on an issue with VDI sessions not completing the provisioning process. I was fortunate to get some time with Ringcube and Citrix engineers to work out this issue.
Here is the issue – I’m using Xendesktop 5.6 MCS (machine creation services) and vSphere 5 u1 for my hosts. The VM’s provision just fine, but are stuck in a “preparing” state. That tells me the VDA is not communicating with the DDCs. Looking at one of my provisioned desktops, I have to log on as a local admin. The AD account is correctly assigned to the VM, so that is not the problem. After first logging in, I notice that the NIC shows disconnected. Also looking in the device manager, everything is gone! How can the device manager be blank? One work around I found was to log into the provisioned VM in “safe mode with networking”, after a reboot everything seems to work correctly. This was not a great solution if I have hundreds or thousands of VM’s deployed in Citrix.
We narrowed down the registry key that was giving us the issue during the provisioning process. It was HKEYLM-System-CurrentControlSet-Enum-LPTENUM.
It was coming down to a device driver that would not load properly because it was missing from the base VM. This LPTENUM reg key was empty in my base VM. I began to step backwards and think about any customization I did on my base VM. One thing I always do before a VM is built is configure the BIOS to disable floppy drives and LPT ports. Why? I disabled anything that may cause unneeded CPU cycles on the vSphere hosts.
The disabled LTP port in the BIOS is what was causing the issue. The LPT port corresponds to the Windows registry entry mentioned above. I was able to disable the floppy controllers with no problem. I also had to remove “name=%SYSTEMROOT%system32catroot2” from C:ProgramDataCitrixpersonal vDiskconfigfolder_rules.txt.
One other thing you will notice when provisioning machines is that the VMware properties of virtual machines do not get replicated to new machines. Settings like “hot add” or “vmware tools upgrade” do not get replicated to provisioned virtual machines. This is also true for BIOS settings, everything is set to default in newly provisioned VMs. So be careful when building your base virtual machines for Citrix MCS with personal virtual disks.
This was a great first experience at EMC world. Of course, everything on the trade show floor is pretty EMC heavy, I think they took a good 40% of the floor. EMC does have a pretty large product line. Compared to VMWorld, you get to see a wide mix of technologies, just not any other hypervisors or VDI brokers. I wonder how the other vCenter management tools like Quest vFoglight and BMC will be treated at VMWorld this year now that VCOps is released? It's all business, you won't see 3par or IBM strage at EMC world.
At EMC world it was nice to see a small presence from Citrix, Wyse and Riverbed. Ohh, and thank you to Riverbed for the Apple TV. I have used they're products before (RSP and Cascade), they are defiantly worth checking out. A big thanks to Unique Digital from houston for taking us out.
Overall I felt there were plenty of great sessions and labs going on at the event. I really didn't care of the Birds of a Feather gathering, it turned into more of a "let's beat up the speaker" instead of having a good discussion. The BOAF sessions I attended were pretty packed, i was hoping for 20 to 30 in a session but they packed in @ 200.
Will I return next year? I am a pretty big EMC fan, but I like to see more diversity. I mainly attended for the customer awards we received for our datadomain and Avamar implementations from the EMC BRS group. It was a pleasure to get to meet BJ Jenkins in person. He is a president of the BRS division for EMC. That being said, I think next year I will attend Interop or Synergy next year. VMWorld is fun and there is a ton of good events + sessions to attend, but i just don't care to visit San Fran twice in a year.
I will admit that it is pretty tough making it to early sessions here in Vegas! I did manage to catch Joseph Dieckhans' session on Troubleshooting Storage performance with vSphere 5. He had a lot of essential information that all admins should be aware of when troubleshooting storage. For years I have always used esxtop to troubleshoot issues and that is still the case today. Joe did show a nice slide of vCenter ops that can give you the same detailed information. Vscsi stats is also a great tool for troubleshooting storage. Joe's presentation gave a great overview of kavg = gavg – davg for looking into latency issues. These are values read in esxtop. There are also circumstances where you may need to dive into the OS of the VM to tune performance when nothing is found in the VMWare or hardware stack.
I'm excited to be a part of EMC World 2012 this year. After last years VMWorld, I had to see how EMC does in Vegas! If you have not built your agenda yet for the conference, please make sure to check out everything in the catalog. There are a ton of interesting sessions going on. Personally, I can't wait to attend the BOAF (birds of a feather) discussion boards. It is also great to be a part of the customer appreciation events this year. I will also be mentioned in the keynote presentation from Pat Gelsinger for backup and recovery with Avamar+Data Domain+ Networker + Data ProtectionAdvisor. Make sure to stop by and see how we've increased our backup performance by 232% and eliminated tape from our backup architecture.
It will also be great to hang out in the bloggers lounge and meet some of my fellow bloggers. Make sure to visit these sites to keep up to on all the great events going on at EMC World 2012. I have attended the past 4 VMWorlds, but this will be my first EMC World. I hope to network with new friends and share success stories from the field.
I've been working on setting up a POC for XenDesktop 5.6 and provisioning machines with MCS (machine creation service). I ran into a problem where the machines I provisioned were trying to PXE boot, as if they were configured to reach my Citrix PVS server. Here is the setup: I have a Win7x64 template in VMware, I have multiple DDC's going through an F5 load balancer. I have both PVS and MCS in the same environment. There is a rule in MCS that the C drive for the template must be on "hard disk 1" within VMware in order for MCS to provision machines correctly. I did have this setup correctly.
Here is the problem I found. My template was using the new "EFI" bios instead of the traditional BIOS. EFI is the next generation BIOS in vSphere 5 for version 8 virtual machines. Since the hard drive device ID is handled differently in EFI vs BIOS, MCS in citrix did not know what device was the C drive upon boot, causing the next device in line (the NIC) to PXE boot. After I built another template with the traditional BIOS (still on version 8 hardware), MCS was able to correctly provision machines.
For an overview of the vSphere 5 EFI (extendable firmware interface) bios, check out this YouTube video.