WilliamLam.com

OVFTool and VMware Cloud on AWS

by // 1 Comment

Recently, I had noticed a number of questions that have come up regarding the use of OVFTool with the VMware Cloud on AWS (VMC) service. I had a chance to take a look at this last Friday and I can confirm that customers can indeed use this tool to import/export VMs into VMC whether they are from a vSphere/vCloud Director-based environment or simply OVF/OVAs you have on your desktop. Outlined below are the requirements and steps that you must have setup before you can use OVFTool with VMC. In addition, I have also include an OVFTool command snippet which you can use and adapt in your own environment.

Requirements:

  1. You must setup VPN connection between your onPrem environment and the Management Gateway on VMC (direct internet access to ESXi is not supported)
  2. Configure the VMC Firewall to allow access between your onPrem and VMC's ESXi host on port 443 (data transfer occurs at ESXi host level)
  3. Specify the Workload VM Folder as a target
  4. Specify the Compute-ResourcePool Resource Pool as a target
  5. Specify the WorkloadDatastore Datastore as a target

Instructions:

Step 1 - Create a Management VPN connection, please see the official documentation here for more details.

Step 2 - Create a two new Firewall Rules that allow traffic from your onPrem environment to both vCenter Server and ESXi host on port 443. vCenter Server will obviously be used for UI/API access and for ESXi, this is where the data traffic transfer will take place.


Step 3 - Construct your OVFTool command-line arguments and ensure you are using the VM Folder "Workloads", Resource Pool "Compute-ResourcePool" and Datastore "WorkloadDatastore" as your target destination since the CloudAdmin user will have restrictive privileges within VMC.

Here is an example command to upload an OVA from my desktop to the VMC vCenter Server:

ovftool.exe `
--acceptAllEulas `
--name=William-To-The-Cloud `
--datastore=WorkloadDatastore `
--net:None=sddc-cgw-network-1 `
--vmFolder=Workloads `
C:\Users\primp\desktop\William.ova `
'vi://*protected email*:*protected email*/SDDC-Datacenter/host/Cluster-1/Resources/Compute-ResourcePool/'

Note: OVFTool also supports the ability to specify a VM that is residing in your vSphere environment as a source, so you do not have to export it locally to your desktop and you can directly transfer it (your client desktop acting as a proxy) to VMC.

Here is the output from running the above command:


Once the upload has completed, you should see your new VM appear in your vSphere Inventory

 

Quick Tip - OVFTool 4.3 now supports vCPU & memory customization during deployment

by // 3 Comments

In addition to adding vSphere 6.7 support and a few security enhancements (more details in the release notes), the latest OVFTool 4.3 release has also been enhanced to support customizing either vCPU and/or Memory from the default configurations when deploying an OVF/OVA.

Historically, it was only possible to modify these values if you were deploying to a vCloud Director endpoint using either --numberOfCpus or --memorySize. When deploying to a vSphere endpoint, these settings were not applicable and users would need to perform an additional operation calling into the vSphere API using whatever automation tool of choice to reconfigure the VM after deployment. It was not the end of the world but also not ideal if you simply wanted to make a minor modification to the default OVF/OVA you were deploying. I definitely ran into this a few times where having this functionality would have been very useful and I know number of customers have also shared simliar feedback in the past.

I had asked whether it was possible to support this use case and it looks like we already had an internal feature request added to the OVFTool backlog and with some additional customer feedback, we were also able to get this enhancement added to the latest release.

The existing --numberOfCpus and --memorySize accepts a VM Identifier (usually the name) followed by the value, for example

--numberOfCpus:Foo=4

The VM Identifier is to help with vApp deployments where you may have an OVF/OVA which is composed of multiple VMs of which you would like to customize each VM with different values. To ensure we do not break backwards compatibility, this pattern has also been extended when deploying to a vSphere endpoint. Having said that, most customers that I have talked to who use OVFTool generally deploy an OVF/OVA that is comprised of a single VM. In this case, rather than specifying the name of the VM again which is derived from --name property, you can simply use the wildcard asterisk (*) to simply apply it to all VMs within the OVF/OVA.

Here is an example of deploying a PhotonOS OVA which is configured with a default of 1 vCPU and 2GB memory and as part of our deployment using OVFTool, we will increase both vCPU to 2 and memory to 4GB:

ovftool --acceptAllEulas --name=Foo --numberOfCpus:'*'=2 --memorySize:'*'=4096 photon-custom-hw11-2.0-304b817.ova 'vi://*protected email*@192.168.30.200/VSAN-Datacenter/host/VSAN-Cluster'

Workarounds for deploying PhotonOS 2.0 on vSphere, Fusion & Workstation

by // 2 Comments

PhotonOS 2.0 was just released last week and it includes a number of exciting new enhancements which you can read more about here. Over the last few days, I had noticed quite a few folks having issues deploying the latest PhotonOS OVA, including myself. I figure I would share the current workarounds after reaching out to the PhotonOS team and seeing the number of questions both internally and externally.

Deploying PhotonOS 2.0 on vSphere

If you are deploying the latest OVA using either the vSphere Web (Flex/H5) Client on vCenter Server or the ESXi Embedded Host Client on ESXi, you will notice that the import fails with the following error message:

The specified object /photon-custom-hw13-2.0-304b817/nvram could not be found.


This apparently is a known issue with the vSphere Web/H5 Client bug with exported vHW13 Virtual Machines. As I understand it, the actual fix did not make it in the latest vSphere 6.5 Update 1 release, but it should be available in a future update. After reporting this issue to the PhotonOS team as I ran into this myself, the team quickly re-spun the vHW11 OVA (since that image also had a different issue) which can now be imported into a vSphere environment using any of the UI-based Clients and/or CLIs. For now, the workaround is to download PhotonOS 2.0 "OVA with virtual hardware v11" if you are using vSphere OR you can install PhotonOS using the ISO.

Deploying PhotonOS 2.0 to Fusion/Workstation

UPDATE (11/08/17) - The PhotonOS team just published an additional OVA specifically for Fusion/Workstation which uses LSI Logic storage adapter as PVSCSI is currently not supported today. You can easily import latest PhotonOS 2.0 without needing to tweak the OVF as mentioned in the steps below, simply download the OVA with virtual hardware v11(Workstation and Fusion) and import normally via UI or CLI.

If you are deploying either of the vHW11 or vHW13 OVA to Fusion/Workstation, you see the following error message:

Invalid target disk adapter type: pvscsi


The reason for this issue is that neither Fusion/Workstation currently support the PVSCSI storage adapter type which the latest PhotonOS OVA uses. In the meantime, a workaround is to edit the OVA to use the LSI Logic adapter instead of the PVSCSI. Below are the steps to convert the OVA to OVF and then apply the single line change.

Step 1 - Use OVFTool (included with both Fusion/Workstation) to convert the OVA to an OVF which will allow us to edit the file. To do so, run the following command:

ovftool --allowExtraConfig photon-custom-hw13-2.0-304b817.ova photon-custom-hw13-2.0-304b817.ovf

Step 2 - Open the photon-custom-hw13-2.0-304b817.ovf using a text editor like Visual Studio Code or VI and update the following line from:

<rasd:ResourceSubType>VirtualSCSI</rasd:ResourceSubType>

to

<rasd:ResourceSubType>lsilogic</rasd:ResourceSubType>

and save the change.

Step 3 - Delete the OVF manifest file named photon-custom-hw13-2.0-304b817.mf since the contents of the file has been updated

Step 4 - You can now import the modified OVF. If you wish to get back the OVA, you can just re-run Step 1 and use the .ova extension to get back a single file

Upgrading from Photon 1.x to 2.0

I also noticed several folks were asking about upgrading from Photon 1.0 to 2.0, you can find the instructions below:

Step 1 - You may need to run the following if you have not done so in awhile:

tdnf distro-sync

Step 2 - Install the PhotonOS upgrade package by running the following command:

tdnf install photon-upgrade

Step 3 - Run the PhotonOS upgrade script and answer 'Y' to start the upgrade:

photon-upgrade.sh