How To: Create A VMFS5 Datastore On A USB Drive
Create A VMFS5 Datastore On A USB Drive
Ever wondered if it was possible to use a USB Drive as a VMFS5 datastore in VMware vSphere 6.0? I sure know that I have! Not that I would like to run any VM’s on said datastore, as I’m sure performance would not be optimal, but instead to test its functionality and use it for storing ESXi host logs for example. Well, I ran into an issue today where I needed to unmount all of my NFS mounts on ESXi 6.0 U2 in order to recreate some of the volumes before remounting them. The problem was that I was unable to unmount one of my volumes because it was bound to the ESXi host for scratch logs. As I didn’t have a spare drive of any sort to attach to my host so that I could reconfigure the location for scratch logs, I began tinkering with the idea of using a small USB drive as a temporary datastore for these logs.
After doing a little research, I came across a post from Florian Grehl aka @virten showing exactly how to accomplish this so I figured I’d share the process of doing so. Keep in mind that this should only be used for testing purposes and should not be used in production environments. This is unsupported by VMware. Here we go!
Ensure that the USB device is unplugged from the ESXi host then begin by connecting to your ESXi host and stopping the USB arbitrator service. This service is responsible for allowing USB device passthrough from an ESXi host to a virtual machine, so keep in mind that you will no longer be able to pass through USB devices to VM’s until this is restarted. (Note: restarting service after creating and mounting USB datastore will break connectivity and recognition of the USB datastore). To stop the service, run the following command:
/etc/init.d/usbarbitrator stop
Optionally, if you’d like to permanently disable the service so it persists thru reboots, run the following command:
chkconfig usbarbitrator off
Plug the USB drive into your ESXi host. For the purposes of this tutorial, I am using a small Lexar 8GB USB device. If you navigate to the storage devices section on your host, you should now see the connected USB device is recognized by the hypervisor. Make note of the device identifier number (mpx.vmhbaXX) for this device.
vSphere Client:
vSphere Web Client:
vSphere HTML5 Web Client:
You can also list the device information to determine the identifier by running the following command:
ls /dev/disks/
As we can see, my identifier is mpx.vmhba40:C0:T0:L0 for this device which also matches the identifier from the GUI pics above. Note: The other USB (mpx.vmhba32:C0:T0:L0) is a separate USB where ESXi is installed on.
Next, we need to create a GPT (GUID Partition Table) label on the device. To do so, run the following command using the correct identifier for the drive. In my case, I will run with mpx.vmhba40 for all of the following commands. Be sure to change this to your correct ID.
partedUtil mklabel /dev/disks/mpx.vmhba40\:C0\:T0\:L0 gpt
Now run the following command to get the partition table information.
partedUtil getptbl /dev/disks/mpx.vmhba40\:C0\:T0\:L0
This returned the following output for me…
gpt 973 255 63 15634432
Next, we need to create a partition in which you will need to know the start sector and end sector which all depend on the size of the device drive and GUID. As an FYI…
- The start sector is always 2048
- The GUID for VMFS is always AA31E02A400F11DB9590000C2911D1B8
- The end sector is calculated using the values obtained by running the previous command.
- Formula: 973 x 255 x 63 – 1 = 15631244
We can also run the following command to calculate the end sector value. This should return an identical value that matches the previous calculation.
eval expr $(partedUtil getptbl /dev/disks/mpx.vmhba40\:C0\:T0\:L0 | tail -1 | awk '{print $1 " \\* " $2 " \\* " $3}') - 1
If everything has gone smoothly so far, you should be ready to create the VMFS partition. Run the following command, ensuring to replace the identifier and end sector values with your own.
partedUtil setptbl /dev/disks/mpx.vmhba40\:C0\:T0\:L0 gpt "1 2048 15631244 AA31E02A400F11DB9590000C2911D1B8 0"
Lastly, we need to format the partition with VMFS using vmkfstools. Do so by running the following (Note: “” in the command below can be any name you like so feel free to use a different name for your datastore):
vmkfstools -C vmfs5 -S USB_Datastore /dev/disks/mpx.vmhba40\:C0\:T0\:L0:1
Sit tight…wait about one minute…and…voila!
After a quick rescan/refresh you should now have and see your mounted VMFS5 USB Datastore!
vSphere Client:
vSphere Web Client:
vSphere HTML5 Web Client:
After I changed the “Syslog” configuration for my scratch logs to use this new datastore, I was finally able to unmount my NFS datastores. I hope this helps so please feel free to comment below.
Shoutout to Florian Grehl for his wonderful post!
Cheers!
-virtualex-
Pingbacks:
USB Devices as VMFS Datastore in vSphere ESXi 6.0
PernixData FVP Freedom Woes With Missing Supermicro System UUID
PernixData FVP Freedom Woes With Missing System UUID
Recently, I’ve been wanting to give PernixData FVP Freedom a run in my HomeLab Datacenter to better familiarize myself with the product and see how much of a performance improvement I’d get if any at all. I’ve heard from so many people how much they love the product so I figured “why not”?
For those who are not familiar with PernixData FVP, it accelerates Storage and Virtual Machines by moving read and write operations to the server tier, instead of the storage tier, using Flash or RAM to ensure the fastest VM performance. This, in turn, reduces VM latency by a claimed 10x and overall SAN utilization by over 80%.
To start off, I visited the PernixData website and went ahead to register for the free FVP Freedom product. A short time later I received an email and obtained my download and license key information, along with all the documentation needed to get it up and running. I installed the ESXi host VIBs and opted to deploy the .ova appliance version so that the deployment would be a piece of cake. Once I got the product up and running, I logged into the Management appliance and attempted to configure my cluster and add resources, but for some reason, none of my hosts’ were showing up. I kept getting the “No PernixData compatible hosts have been detected in the cluster“, and only (1) of my (5) hosts was detected but it was not part of the cluster that I was configuring yet.
This is where I ran into a snag that took quite a bit of time to research and find a fix. Luckily, another blogger by the moniker “vWilmo” who’d experienced this same issue and described how to fix it, so I figured I’d write a similar entry for my own reference, and to help others who may frequent my blog. I will also be sharing his link at the bottom of this page.
Ultimately, the issue stemmed from the fact that Supermicro did not generate any system UUIDs for my boards and FVP needs them to detect the hosts to use as resources. KB 1006250 references the situation of an ESXi host not having a unique UUID but did not offer a solution other than to contact the manufacturer (which I did via email and am still awaiting a reply). To confirm this, I ran a script I found online called Get-VMHostUUID to pull the UUID’s from all of my hosts connected to my vCenter server. Upon review, it only returned a value for my “white box” host, and returned all “zero” values for my remaining (4) Supermicro systems. I also ran prnxcli via SSH connection to my host which returned an error as well.
As my Supermicro systems run an AMI (American Megatrends) BIOS, there is a BIOS utility that can be used to generate a new UUID for the system which can be found here. Download this file and extract the contents. The file we need to use is named AMIDEDOS.exe, so I took this file and placed it on a DOS formatted USB drive that I had created with Rufus back when I needed to flash my BIOS and upgrade my Intel NIC firmware.
Insert and boot into the USB, then navigate to the directory that houses the file mentioned above.
Enter the following command:
AMIDEDOS.exe /su auto
If successful, this will generate a new system UUID for you and you should then receive an output like this:
Reboot your host, SSH into it and run the prnxcli command. If it runs successfully you should see an output like this:
After I completed this process on all of my (4) impacted hosts, I ran the Get-VMHostUUID script again and was happy to see that I now had a valid UUID for each host which match each prnxcli output.
Upon logging into the vCenter Web Client, I noticed that there is now a PernixData plugin icon in the vCenter Web Client interface which can be selected to launch the PernixData Management Console or access the FVP dashboard from within the Web Client.
Lastly, I logged into the PernixData FVP Management console again, and I was now able to create my cluster and assist hosts as resources. The only caveat is that there is a single-cluster limitation with the FVP Freedom version license, so if have all of your hosts in a single cluster then you are good. Unfortunately for me, I have (3) clusters so I need to pick which one I want to use FVP with. I decided to use my Management Cluster since that houses the majority of my VMs at the moment.
After letting it work its magic for a few hours, I noticed that the VM latency had reduced drastically to an average ~ +/- 2.0 ms and overall performance was great! I must say that I am really impressed, satisfied, and glad that I gave this program a shot!
Well, I hope that you have found this useful, thanks for stopping by!
Special thanks to Geoff Wilmington aka @vWilmo for helping me to solve this as I am lucky I found your post. Another shout out to Andy Daniel aka @vNephologist from PernixData for his willingness to communicate with and try to assist me with this problem.
Pingbacks:
