locked
Slow PXE boot and Task Sequence execution after migrating to VM RRS feed

  • Question

  • I recently migrated my MDT/WDS server from a physical machine to a VMware virtual machine. Prior to migrating it, my PXE TFTP transfers took less than 20 seconds and they are currently taking a few minutes. Nothing in my environment changed other than migrating to the virtual. Additionally, after migrating to VM my boot.wim loads and boots but does not run the start of my task sequence. This happens intermittently, sometimes a reboot or two fixes the issue and the task sequence runs without any issues. I have attempted to change my max TFTP block sizes to 16784 but this had no effect. I am only running these deployments in unicast with a max of three or four machines pulling images. Today, for instance I have two machines on my bench. One imaged fine, the other wont contact the Deployment server. It does not even register with the MDT monitoring server. Not sure what else can be done. I have not done any packet caps but given that nothing in my network changed I am not sure that would yield any results. 
    Friday, December 9, 2016 3:17 PM

All replies

  • I had a similar problem, my suggestion would be to check the NICs on the Virtual Machine. I found that when I migrated from a physical machine (Where I had multiple NICs) to a Virtual machine, the NICs were incorrect since it was migrated.  I had to many NICs after migration. 
    Friday, December 9, 2016 3:23 PM
  • I had a similar problem, my suggestion would be to check the NICs on the Virtual Machine. I found that when I migrated from a physical machine (Where I had multiple NICs) to a Virtual machine, the NICs were incorrect since it was migrated.  I had to many NICs after migration. 
    Thanks for the reply, I glanced at the NIC and everything seems to be fine. I did a complete rebuild of the MDT share on a clean install of 2012 R2 that was installed on the virtual. I did fail to mention that but I am wondering if it has something to do with the VMware switch that machine is connected to.  
    Friday, December 9, 2016 4:27 PM
  • Sounds like there is a new bottleneck somewhere after the switch to virtual. Some causes I can think of off the top of my head:

    1. Different physical switch topology or settings for the virtual environment (Jumbo frames, different switch capacities)
    2. Different virtual switch settings or limits
    3. Network saturation on the virtual host machine
    4. Slow reads from the datastore the MDT server is on

    To isolate the issue try PXE booting a client VM that is on the same host and network as the MDT server to see if the issue is local to the vmware host (virtual switch problems, slow datastore reads, or other resource contention).

    Friday, December 9, 2016 6:11 PM
  • I had an opportunity to re-image one of my reference machines and the problem does impact the load times of a VM machine on the same host connected to the VM switch. 
    Friday, December 9, 2016 9:17 PM