Tag Archives: VMware

Trying Herald OS on VMware Fusion – Giving an AI Agent Its Own VM

Recently I came across Herald OS, and it immediately earned a place on the list of things I wanted to try. An operating system built around an AI agent, with the agent able to work across the machine? That sounds like something that belongs in a lab for now and im interested in what this means for the OS we use every day and if this would improve productivity.

I already had VMware Fusion installed on my Mac, so that seemed like a good place to start. There was an ARM64 disk available, a bit of conversion to do, and one missing library waiting to interrupt the afternoon.

I have now rebuilt it from the original image and checked that it survives a proper shutdown and cold boot. Here is the route that worked, including the fix, so you can spend more time exploring it.

What is Herald OS?

Herald OS puts Hermes Agent at the centre of the desktop. Its interface brings together conversations, files, memory, missions and scheduled work, with permission-controlled tools for acting on the machine. It is an independent project by Luke The Dev, using Nous Research’s Hermes Agent.

The Linux version uses Fedora underneath, the niri compositor to arrange windows, and Herald’s own shell for the desktop panels. Its bridge connects Hermes to system tools. There is also a macOS app, but I wanted the Linux environment in a VM. The Linux documentation explains how those pieces fit together.

It is still an alpha. That made a VM an easy choice: give it its own disk and a separate environment to experiment in.

What I used

This walkthrough uses the 0.1.0-alpha.4 aarch64 release image on an Apple Silicon Mac. I tested it with Fusion 13.5.1. Treat that as the version I ran, rather than a minimum-version or host-support claim.

SettingMy configuration
Guest architectureARM64 / aarch64
Virtual CPUs4
Memory8 GB
FirmwareUEFI, Secure Boot disabled
Disk controllerNVMe
NetworkNAT, vmxnet3
Graphics3D acceleration enabled

The release disk has a virtual capacity of approximately 41.5 GiB. Leave room on the Mac for the compressed download, the unpacked QCOW2, the converted VMDK and the applications installed during first boot.

The project’s documented Apple Silicon VM route uses Apple’s Virtualization framework through vfkit. The Fusion conversion here is the adaptation I tested. Use the aarch64 disk for this walkthrough; the x86_64 installer is for a different architecture.

First, grab the tools and helper scripts

On the Mac, install the conversion tools and a current Python using Homebrew:

brew install python qemu zstd
python3 --version

The helpers need Python 3.11 or later. They assume Homebrew is installed under /opt/homebrew and Fusion is in /Applications.

I have included the scripts used for the build:

Download the Fusion helper scripts

Unzip that archive into Downloads. It contains a herald-fusion-helpers folder with:

  • create_vm.py — verifies the release image, converts the disk and writes the Fusion configuration.
  • serial_console.py — connects to a temporary local serial console.
  • finish-setup.sh — applies the missing-library fix inside Linux and finishes Hermes setup.
  • normal_boot.py — removes the temporary serial-console configuration and starts the completed VM.

There is a README in there too. Have a look through the scripts before running them; they are short enough to see what each one changes. The VM creation script refuses to overwrite an existing VM.

Download the release disk

Run these commands on the Mac. They deliberately use the release tested here:

mkdir -p "$HOME/Virtual Machines.localized/Herald OS Lab"
cd "$HOME/Virtual Machines.localized/Herald OS Lab"base=herald-os-0.1.0-alpha.4-aarch64.qcow2
release=https://github.com/iamlukethedev/Herald-OS/releases/download/v0.1.0-alpha.4 for file in "$base.zst.part0" "$base.zst.part1" "$base.zst.sha256"; do
curl -fL --retry 3 -C - "$release/$file" -o "$file"
done
cat "$base.zst.part0" "$base.zst.part1" > "$base.zst"
shasum -a 256 -c "$base.zst.sha256"
zstd -d "$base.zst" -o "$base"

The compressed disk comes in two parts. Join them in order, check the checksum, then unpack it. If you already have an unpacked disk, do not decompress over it.

Convert it and create the Fusion VM

Back on the Mac, switch to the extracted helper directory:

cd "$HOME/Downloads/herald-fusion-helpers"python3 create_vm.py 
"$HOME/Virtual Machines.localized/Herald OS Lab"
--start --serial-console

The script checks the archive checksum, confirms the QCOW2 matches the archive’s contents, and runs qemu-img check. It then converts the disk into a sparse VMDK and compares the two disks before starting Fusion.

The conversion itself uses:

qemu-img convert -f qcow2 -O vmdk \
  -o subformat=monolithicSparse source.qcow2 destination.vmdk

That line explains the conversion; the helper supplies the actual filenames and builds the .vmwarevm bundle around it.

The --serial-console option gives us a way into Linux while the first desktop is starting. It uses a local socket on the Mac and we remove it once setup is complete.

The bit that caught me: libatomic

The first boot downloads Hermes and its dependencies. In my clean test, that got as far as Node and then failed with:

error while loading shared libraries: libatomic.so.1:
cannot open shared object file: No such file or directory

Node needs libatomic. The installer tried to install it with dnf, but this Herald image uses a read-only bootc system. That ordinary package-install attempt failed too.

The repair I used downloads Fedora’s ARM64 RPM, verifies its signature, and extracts the library into a persistent directory under /var/lib. An entry under /etc/ld.so.conf.d tells the loader where to find it. This avoids changing the bootc image or relying on a temporary overlay that disappears at reboot.

This is a local workaround for the release I tested. If a later image includes the library, this particular repair should no longer be needed.

Finish setup through the serial console

First, make the helper available to the VM. In a Mac terminal, from the helper directory:

python3 -m http.server 8766 \
  --bind 172.16.206.1 \
  --directory "$HOME/Downloads/herald-fusion-helpers"

Check that address for your Fusion network. 172.16.206.1 is the Mac-side address of my Fusion NAT network. Yours may differ. Use ifconfig on the Mac to find the bridge address on the same subnet as the VM; the VM’s serial login banner shows its own address. Replace the host address in both the server command above and the download command below.

This server only needs to run during setup. Binding it to the Fusion host address keeps the download on that interface.

In a second Mac terminal:

cd "$HOME/Downloads/herald-fusion-helpers"
python3 serial_console.py

At fedora login:, enter hermes. The default account in this release initially has an empty password. Let the automatic first-boot attempt finish before applying the repair; you can check its state with:

systemctl show herald-os-firstboot.service -p ActiveState

If it is still activating, give it more time. To inspect its recent output without leaving the shell, run journalctl -u herald-os-firstboot.service -n 20 --no-pager.

Now run the following inside the VM, at its Linux shell prompt:

curl -fsS http://172.16.206.1:8766/finish-setup.sh \
  -o /tmp/finish-herald-setup.shbash /tmp/finish-herald-setup.sh

The helper verifies the Fedora RPM, installs the library, reruns Hermes setup, enables Herald’s bridge and restarts the graphical session. It prints the Hermes checkout it used, which is useful when comparing results later.

The library registration at the heart of the fix looks like this:

echo /var/lib/herald-local/lib | \
  sudo tee /etc/ld.so.conf.d/herald-local.conf
sudo ldconfig

The complete script in the download also handles fetching, verifying and extracting the package. Those steps need to happen before registering the directory.

Prove it is running

A desktop appearing is a good start. I also wanted to see the library resolve and the Hermes backend start successfully.

Run these inside the VM:

ldconfig -p | grep libatomic
~/.local/bin/hermes --version
grep HERMES_BACKEND_READY \
  ~/.local/state/herald-os/shell.log | tail -1

You should see libatomic.so.1 resolve to /var/lib/herald-local/lib, a Hermes version, and a HERMES_BACKEND_READY message. The port number is allocated when Hermes starts, so it can differ from the screenshot.

The rebuilt VM showing libatomic resolved, the installed Hermes version and HERMES_BACKEND_READY.
Actual checks from the rebuilt VM after a cold boot: the library is present and Hermes has started its backend.

My clean run installed Hermes v0.22.5. There is one detail to keep in mind here: pinning the OS disk does not pin everything the first-boot installer downloads. Its Hermes checkout can move upstream.

If the desktop is blank after the failed installation, restart the graphical session after completing the repair:

sudo systemctl restart greetd
sudo chvt 1

The helper does this already. It ends your current graphical session, so use it during this initial setup before you have work open in the VM.

Give it a proper cold boot

Inside the VM:

sudo systemctl poweroff

Once Fusion shows the VM powered off, run this on the Mac:

cd "$HOME/Downloads/herald-fusion-helpers"
python3 normal_boot.py \
  "$HOME/Virtual Machines.localized/Herald OS Lab"

That switches the temporary serial console back to a log file and boots the VM normally. Stop the temporary HTTP server with Ctrl-C in its Mac terminal. You can close the serial-console client as well.

This is the check that mattered to me: after powering it off and starting it again, the welcome screen returned and Hermes started. The repair survived the reboot.

Add VMware Tools

One more thing to check before settling in: VMware Tools. Fusion initially reported them as installed, but checking inside this release of Herald showed that the packages and service were missing. The guest checks were the useful ones here.

Herald uses an immutable Fedora image, so I installed Fedora’s open-vm-tools packages with rpm-ostree. Run these commands inside the VM, in its terminal, and let the installation finish before rebooting:

sudo rpm-ostree install open-vm-tools open-vm-tools-desktop
sudo systemctl reboot

After the desktop returns, open a terminal and check:

rpm -q open-vm-tools open-vm-tools-desktop
systemctl is-active vmtoolsd
systemctl is-enabled vmtoolsd
vmware-toolbox-cmd -v

My VM reported both packages as 13.1.0-2.fc44.aarch64, the vmtoolsd service as active and enabled, and the toolbox version as 13.1.0.0.25218885. This confirmed the tools started automatically after reboot.

The installed VMware Tools packages, active and enabled vmtoolsd service, and toolbox version inside Herald OS.
VMware Tools verified inside the guest after reboot. Fusion also reported the tools as running and retrieved the guest IP address.

I installed the desktop package too. Clipboard sharing, drag and drop, and automatic resizing have not been verified in Herald’s niri/Wayland session; these service checks establish that the system tools are running.

There is an update detail to keep with these instructions: installing the packages creates a local package layer. While that layer is present, use sudo rpm-ostree upgrade, followed by a reboot, for OS updates. bootc upgrade does not support those local package modifications. The bootc documentation explains the interaction, and the rpm-ostree handbook covers package layering and upgrades.

Now make it yours

Click It is for me, choose your name, set a VM password and configure a model provider. The Herald setup documentation covers signing in and selecting a provider.

Once you have signed in, a useful first request would be:

Show me this machine's CPU, memory and disk usage.

I verified the VM, desktop and Hermes backend for this post. I left provider authentication to the personal setup stage, so that example is your next test, rather than a conversation I have already run.

What interests me now is trying the workflow: give an agent an environment of its own, see which everyday tasks become easier, and keep an eye on what it asks permission to change. The VM is ready. Time to have a play.

New Year, New Look vCheck

Its been a while since I blogged about vCheck but that doesn’t mean there has been lots of work ongoing with the project, in fact there has been multiple releases and hundreds of pull requests with great new plugins checking for even more issues with your vCenter and lots of bug fixes, in fact this project and the community updating and using it is awesome, here are some the current stats:

  • 954 Commits
  • 64 Contributors (Thanks to all here)
  • 66 Open Issues and 269 Closed Issues

Whats more, one of the most recent releases 6.25 here includes a great new look, the new look is based upon VMwares Clarity framework and personally I think it looks fantastic as you can see below:

vCheck Clarity

Also available (currently in the dev branch) is the brand new dark version of the theme as seen below:

vCheck Clarity Dark

Download for free now!

Download the latest version here:  https://github.com/alanrenouf/vCheck-vSphere/archive/master.zip

And the dev branch version here: https://github.com/alanrenouf/vCheck-vSphere/archive/dev.zip

 

Announcing VMware VR Datacenter Experience

It gives me great pleasure to announce a new open source project from VMware called VMware VR Datacenter Experience, this is an immersive experience that allows you to put on your virtual reality headset and take yourself into a virtual management environment where you will be able to control your vSphere Environment with the wave of a hand or click of a button.

Details

The VMware VR Datacenter Experience allows you to use a HTC Vive headset to enter a virtual datacenter where a virtual representation of each of your hosts is created by integration with the vSphere 6.5 REST API, once these are created the VM’s are dropped onto these hosts, one for each VM currently registered with the host.  In this environment you are able to teleport yourself to each of these hosts and interact with the VMs.

You can pick them up to find out more information about their configuration:

VM Details

 

 

 

 

 

You can even drop them into a trashcan to remove them from your datacenter:

VM Trash

 

 

 

 

 

 

We even extended it to work with VMware Cloud on AWS (VMC) where you could literally throw (vMotion) your VM’s from your On-Premises datacenter into VMC.

Showing you the 2D pictures of this virtual reality environment really does not do it justice, you need to try it to believe it!

Background

At VMware we have internal Hackathon’s where we are able to take a couple of days away from our day job to work on projects which we think are interesting and could provide valuable to VMware, this is a great initiative that fosters

innovation, personally I love working for a company that provides this level of free thinking!  Together with William Lam, Yateendra Kulkarni, Vinod Pai and Steve Trefethen, I put a team together and we set out to work on what a VMware Datacenter would look like in Virtual Reality.

I should mention at this point that none of us had developed for VR before, in fact we had never even used the IDE (Unity) before, that was fun!  After a couple of days we were pleased when we managed to produce a single host and VMs on that host all by calling the vSphere 6.5 REST API.

After this and some internal marketing and some spare time hacking away at the project we came up with something pretty cool that later allowed me to realize one of my bucket list items and present on the main VMworld Keynote stage with Pat Gelsinger (VMware CEO).

VMworld Keynote

If you didn’t get a chance to watch the VMworld Barcelona Keynote then please do take 5 minutes out of your time to watch the demo below that shows the full remit of what we expanded this project to work with.

Where Can I Get this?

The initial project is now available and ready for you to download, it has been cut down to its basics so we can build on it and ensure the code we are adding and the functionality is based around user input.

Of course you will need the following:

Check out the project landing page here for more details: https://vmware.github.io/vr-dc-ex/

And if you want to take a look at the code, ask for more features or even contribute back to the project then please checkout the github repo here: https://github.com/vmware/vr-dc-ex

The Full story

At VMworld I had the chance to catch up with Barry Coombs who interviewed my on this complete story, check out the clip below to learn how a small town boy ended up on stage with a Silicon Valley CEO…

Ensuring peak configuration for vSphere with Runecast

Back in the day when I could spend time working on vCheck (I miss those days!) I always wished there was a programmatic way to look at the data in the VMware KB articles and be able to mine the information for potential issues or be aware of enhancements to the configuration I had made with my specific version of vCenter or other VMware products.

A few weeks back, whilst at the Melbourne and Syndey VMUGs in Australia I was introduced to a pretty awesome company and product that sparked my interest.. Runecast (https://www.runecast.biz/)

Recently I managed to find time to install this in my home lab and get started, I have to say, I was impressed.

What is Runecast?

Runecast is a virtual appliance which checks your environment, vCenter, ESXi servers and VMs to ensure they are running at peak performance with no known configuration issues or security issues.

The appliance is updated from the internet or can also be updated via a ISO image attached to the virtual appliance for environments without internet access.  It is updated with information which has been mined from the VMware KB articles and is used to proactively check log and configuration for issues in the virtual environment.

Additionally the information from the security hardening guide or what is now known as the vSphere Security Configuration Guide is also added to the virtual appliance to ensure security configuration is also taken into consideration. Continue reading →

Automating the VSAN HCL with PowerCLI

Recently I was contacted by a customer who needed to be able to update their VSAN Hardware Compatibility List in the VSAN Health Check but was unable to do so via the GUI as their vCenter servers did not have internet access.

This is a common setup as a lot of customers clearly do not want their Server infrastructure having a direct connection to the internet due to strict security requirements. The problem is when the vCenter server needs to update the VSAN HCL database file it requires a connection to the internet to do this. Whats more, this specific customer had several VCs and was getting quite frustrated with the warning that the HCL database had not been updated. Rather than turning this feature off the customer was looking for a way to update the HCL from a computer that had internet access – his desktop.

New in PowerCLI 6.5 (backwards compatible to previous versions) is a cmdlet that will help us achieve this… Update-VSANHCLDatabase, as you can see from the below image this can be run either grabbing the database information directly from the internet or if you add the “FilePath” parameter you can load the database locally.

Continue reading →

Automating the build of your vSphere 6.5 home lab

This year at VMworld SFO and BCN I was involved in organizing a couple of great hackathons with the @VMwareCode guys and William Lam, these were highly successful and I have to say the highlight of both my VMworlds.  The teams were great, the end projects were fantastic and most of all, everyone that attended told me they learned something, this if you ask me was the main objective for the hackathon.

If you attended I do want to extend a huge thanks for joining in, having fun and learning with us.

credc8rusaa8un0

To put the hacakthon together we needed to build up some environments for people to use, William came up with the idea of using Intel NUCs as these were easily transportable and packed a punch for their size, the equipment we purchased is listed below:

Quantity: 2 Crucial 16GB Single DDR4 2133 MT/s (PC4-17000) SODIMM 260-Pin Memory – CT16G4SFD8213
Quantity: 1 Intel NUC Kit NUC6i3SYH BOXNUC6I3SYH Silver/Black
Quantity: 1 Samsung 850 EVO 500GB 2.5-Inch SATA III Internal SSD (MZ-75E500B/AM) (For Capacity)
Quantity: 1 Samsung SM951 128 GB Internal Solid State Drive MZHPV128HDGM-00000 (For Performance)

For the hackathon we needed to build a lot of these units, whilst we did some parts of it manually William and I recently took the time to complete the automated deployment of these units, in fact the script is not specific to these units, it will work on any ESXi host with 2 disks, one for performance and 1 for capacity.  Of course you can also adjust the script to use more disks if you have them!

Once you have ESXi on a USB insert it into the machine and configure it to boot from the USB, after the ESXi machine is on the network you can alter the configuration settings in the start of the script and run the script in the Deployment Script section of this blog to automate the following:

  • Configure VSAN in a single node configuration (Unsupported)
    • Use the smaller SSD for performance
    • Use the larger SSD for Capacity
  • Configure NTP on the ESXi Host
  • Enable SSH on the ESXi Host for debugging
  • Configure the Syslog settings on the ESXi Host
  • Deploy the VCSA on the ESXi Host
  • Enable VSAN Traffic on the Management Network
  • Create a Datacenter
  • Create a Cluster
  • Create a subscribed content library for William Lams Nested ESXi Library
  • Enable Autostart so the VCSA VM starts when the ESXi machine powers on
  • Enable SSH on the VCSA Server

Continue reading →

A Quick Reference of vSphere IDs

How do you identify the components within your virtual environment? You give them names right?! Well to programmatically identify the same components that’s exactly what VMware does but rather than names they give them IDs and guess what, there are a lot of different IDs, so many that people often get mistaken by which ID to use in which circumstances and what they refer to.

After a discussion with one of VMware’s brightest engineers (Thanks Jeff Hu) and a long email thread where he dumped his brains on what these IDs were for, I thought I would blog them and make use of this awesome information.

vCenter IDsSo here goes, here is a list of a number of the different IDs used in the virtual infrastructure and how they are used:

image1. VC Instance UUID (aka serverGuid): This is a GUID that identifies the vCenter server.  This is accessible from the AboutInfo property in the VIM API.  The property name is AboutInfo#instanceUuid.  It’s set at install time and is persisted so the value is durable for a given VC instance.

To access this via PowerCLI you can use the following:

PS C:\> $vcenter = Connect-viserver vcsa-01a.corp.local -User Administrator@vsphere.local -Password VMware1!

PS C:\> $vcenter.InstanceUuid

d322b019-58d4-4d6f-9f8b-d28695a716c0

2. The ESX Host UUID is read by the ESX host from the System Management BIOS (SMBIOS). This UUID is not generated by VMware, it is unique to the hardware and is set in the BIOS by the vendor.

To access this via PowerCLI you can use the following:

(Get-VMHost | Select -first 1).ExtensionData.hardware.systeminfo.uuid

3. VC-VM Instance UUID: This is the vc.uuid property in the VM vmx configuration file and also the instanceUuid property on the VmConfigInfo for a VirtualMachine in the VIM API.  This UUID is used by vCenter to uniquely identify VMs that it is managed.  vCenter assigns this UUID to the VM after it is created.  There are different workflows where a VM may not have a UUID initially or may have a duplicate.  vCenter actively looks over the VM inventory and will automatically patch the VM instance UUID if it sees a duplicate.

The scope of this uniqueness scan is a vCenter instance.  If there are duplicate VC-VM Instance UUIDs, they will not be detected as duplicates unless the VMs are managed by the same vCenter instance.  This value can be changed through the API.

To access this via PowerCLI you can use the following:

PS C:\> (Get-VM | Select -first 1).extensiondata.config.InstanceUUID

502e363d-393c-b0a4-8ede-639db30a625e

4. VM SMBIOS UUID: This is the uuid.bios property in the VM vmx configuration file.  In the VIM API, this is the uuid property on the VmConfigInfo for a VirtualMachine managed object.  This UUID is visible to the guest OS as the SMBIOS UUID.  VMware generally try to avoid changing this UUID since it is often used by applications that manage the guest OS instances.  When the VM is migrated this UUID is maintained.

vCenter does a best effort to try to make sure these UUIDs are unique in an environment.  For operations like clone and deploy that create a new VM, a new UUID is provided to the VM by clearing the uuid.bios from the VM’s vmx file.  If the VM is copied without vCenter’s knowledge, 3rd party tools are advised to do the same thing as vCenter.  A new uuid.bios is generated the next time the VM is powered on.  This UUID is generated by the ESXi server so vCenter does not have to be in the control flow for the uuid.bios to be generated.

To access this via PowerCLI you can use the following:

PS C:\> (Get-VM | Select -first 1).extensiondata.Config.UUID

422ee9f4-6cfd-1fb7-1376-a8ecb8eb8568

5. VM Location ID: This identifier is stored in the VM configuration file as the variable uuid.location.  In the VIM API, the name is locationId also on the VmConfigInfo of the VirtualMachine.  This is one other detail that might be relevant to a 3rd party solution that might be moving the VM without vCenter’s knowledge.  This property is used in conjunction with the VM SMBIOS UUID.  This property is a hash of the VM’s configuration file and some UUID of the ESXi. The purpose of this UUID is to detect out-of-band migrations and copies of the VM.  If vSphere (or Workstation/Fusion) detect that the locationId is no longer valid, it will pop up dialog/question/error asking the user why the locationId is no longer valid.  The user is prompted to answer whether the VM was moved or copied.  The purpose is so that vSphere can determine whether it should keep the existing the SMBIOS UUID or generate a new one.

Recall, I mentioned that vCenter is doing a best effort to ensure the SMBIOS UUID is unique.  But when it detects a potential conflict, vCenter will not automatically patch the UUID since this could break 3rd party and VMware products that rely on the guest visible UUID to correlate the VM with other management entities in the environment.

The way 3rd party solutions can best interoperate with vCenter and it’s attempt to preserve a unique SMBIOS UUID is to be aware of these semantics.  If a 3rd party solution or program is moving the VM, it should clear out the uuid.location (ie. ConfigInfo.locationId) property.  On next power on, a new uuid.location will be generated, binding the VM to ESXi host and the file path where it is running.

Despite this however, there is a theoretical and real possibility that there could be duplicate SMBIOS UUIDs.  Because vCenter is deferring to what the user of the system what the the SMBIOS UUID should be, it will not change the UUID automatically if it detects conflicts.  If there SMBIOS UUIDs that are duplicated, it will not fix them.

To access this via PowerCLI you can use the following:

PS C:\> (Get-VM | Select -first 1).extensiondata.config.LocationId

564d043c-5dab-5413-c4a7-fb0837d354d9

6. VM MoRef: The VM MoRef is a short key used by a vCenter instance to identify the VM.  This is the primary identifier used by the VIM API to refer to the VM.  The other identifiers can generally be used to find the VM, but the VM MoRef is the one that must ultimately be used to get data and issue operations to modify the VM.

In general, the MoRef is the important identifier because that’s what is used in the VC database to associate a VM to its stats, events, and configuration.

In environments where multiple vCenters are used, the MoRef will not necessarily be unique.  As a result, another identifier needs to paired with the VM MoRef or used instead.  In general, vSphere uses a combination of the vCenter instance UUID (ie. the serverGuid) with the MoRef to identify the VM for cross-vCenter scenarios like Linked Mode.

To access this via PowerCLI you can use the following:

PS C:\> (Get-VM | Select -first 1).ExtensionData.Moref.Value

vm-43

More Information

William Lam also has some awesome posts on IDs which you may find useful, please read these for further related information:

Conclusion

A lot of Information and different IDs I know but hopefully this will help you identify which to use and when so that you can get the most out of your Virtual Infrastructure!

Adding a vGPU For a vSphere 6.0 VM via PowerCLI

I had a conversation with a VMware customer the other day and they were asking if there was a way to automate the addition and removal of a vGPU to a VM dynamically on a requested or scheduled basis, VMGuru has a great post here on what exactly a vGPU is, I highly recommend reading it.

They asked for a PowerCLI command or function to configure the VM graphics card, specifically to assign a vGPU resources to the VM, a desktop user’s VM is configured by default with vSGA (to not tie up  vGPU resources) then the user schedules the vGPU resource for a specific time frame using their end user portal, before the scheduled vGPU reservation time frame the user is logged out and the script would kick in. The script would reconfigure the VM settings and assigns a vGPU profile to it, the equivalent of what is being done in the UI as follows:

image

Continue reading →

VMware Flings – The inside story

Recently I was asked to participate in a video where I spoke about the recent Onyx For Web Client fling, this was the winner of the Fling contest held in 2015 and produced and released a while back, more information on the fling and how to use it can be found here, and an informational video can be seen at the end of this post.

As part of the fling I was asked some questions on how it was achieved and what flings take to produce at VMware, I have always liked the idea of flings and fully support a number of flings from VMware both internally to the company and externally, as an addition to the video I thought I would also share the answers to the questions I was asked and give you an insight on what it takes to get a fling out and what it means to the development staff at VMware.  Something I would highly recommend in any company to help create and spread innovation throughout the products.

Onyx For the Web Client was certainly close to my heart, being able to give developers and system administrators a record and show feature that allows them to investigate our APIs and learn more is always a good thing.

Find below some Q&A on the fling:

How did you and your team bring the concept from vision to reality?

When someone from the CTO office calls your phone, you answer! But honestly, we already had seen the value and the tremendous following with the old version of Onyx and were aware of the limitations it had, specifically when moving to the web client.  We had already been looking at a web client version of Onyx so when this item won the fling contest we jumped at the chance to be involved.

As a customer I used the original version of Onyx on a weekly basis, the value of being able to automate your infrastructure is huge and if you can find out how VMware does a particular task and convert that to code it doesn’t take much to include it in your custom automation solution and prevent yet another manual task, after all isn’t this the promise of the SDDC J

Being the Product Manager of the PowerCLI team all I needed to do was share the vision with an eager team member, in this case it was Atanas Atanasov, a forward thinking developer in the PowerCLI team, someone who is keen to make an impact and provide the best solution for our customers.  He took some original work that had already been started as part of an intern project and polished it up, added features and ran through the entire development and internal feedback process to produce an awesome fling!

Innovation is hard to manufacture. How did your team at VMware collaborate to innovate in the context of this project?

Innovation in my mind is coming up with a unique way to solve an important customer issue, this is exactly what we did in this case.

When moving towards an automated datacenter, VMware customers need all the help they can get to quickly move from manual tasks to automated tasks and the most popular method of doing this at the moment is PowerCLI, this fling helps our customers understand exactly what happens “under the covers” and replicate this in their scripts.

How does the VMware leadership team help support innovation within the company? How did this affect the development of the Onyx for the Web Client Fling?

The VMware leadership team were fully supportive.  Flings are often worked on in the developers spare time at VMware and I can tell you as the PM for PowerCLI we do not give the developers any spare time!  The VMware leadership team helped me to bring incentives to the engineer, in this case a fully endorsed trip to VMworld where he could share his hard work with the customers that he has helped in producing this very fling.  Atanas and I were able to stand on the Office of the CTO booth talking about the fling and helping customers with automation. It was great to receive their direct feedback and watch their faces when they saw it in action and how much time it would save them.

What’s the reaction been to the Fling since its release?

WOW! The community for PowerCLI and automation was abuzz with the news of this fling, I lost count of the social media comments and blog posts which called this out as an amazing achievement from VMware and a great help to our customers.   When you have a good solution that customers see the value in it is a great day to be a Product Manager at VMware!

Besides this Fling, what’s your favorite Fling?

There are many amazing flings on the flings site, it’s hard to choose just one.  I guess I would need to stay on the automation theme though.

PowerActions is probably my favorite fling, this is kind of the reverse of “Onyx for the web client”, it gives our customers a way to feed their automation back into the web client and perform custom scripted reports and actions on the items in the web client.  Essentially it allows our customers to customize the GUI and gives them a way to perform automated tasks in their day to day console.  Not only this but also share the wealth of expertise that the senior automation engineers have by allowing them to create custom actions for everyone who uses the web client.

Together with “Onyx for the web client” I believe PowerActions is an amazing solution for the automated Software Defined Datacenter.

More Information

Make sure you check out the VMware RADIUS site here for more information on this and watch the video below:

PowerCLI 6.0 R2 – The most advanced version VMware has ever made

imageThis week VMware released a new version of PowerCLI and this release is the most advanced version ever made.  As the Product Manager for PowerCLI I am particularly proud of this release as we not only add core functionality to the vSphere cmdlets allowing access to report, manage and automate even more of the core infrastructure but yet again we introduce a number of new features allowing PowerCLI to reach even further into the SDDC and automate and integrate more products.

I always like to ask people what they want to see from PowerCLI next and for a long time now people have been talking to me about vROPs and telling me their use cases of being able to use the rich dataset and analytics provided by vROPs to make decisions and take further action with PowerCLI, something as simple as monitoring a web server to know when it is under duress and provision many more based on these statics all the way to being able to take action on storage systems by monitoring the workload over time.  The use cases are endless and now achievable by the latest version of PowerCLI.  More on this to come in a future post but lets just say I am excited to see what the awesome PowerCLI community does here.

There is so much to talk about in this release, I wanted to give you an overview in this post and then dive deeper into some of the key features in the future.

What’s New?

More Module Enhancements

The community made it clear that we had to move to modules, and whilst we are still getting there, in this release we made even more module enhancements, both to the core distribution model of PowerCLI cmdlets and also converting more of our code base from snapins to modules. In this version the License snap-in has been converted to a PowerShell module. In the new release the PowerCLI Modules have also been moved to the System PSModulePath allowing all users of a machine to access them once installed.

vROPS Support

As briefly mentioned above, a new module and cmdlets have been added to this release to allow access to vRealize Operations, access to the entire public API is available from the $global:DefaultOMServers variable using the ExtensionData property and full cmdlets have been included for the most used features, this gives us a good mix of fully functional easy to use cmdlets and yet the access for advanced users to be able to access the entire API through the ExtensionData property to create their own functions to expose anything to automate vROPs.  The following cmdlets were added for using with vROPs:

  • Connection: Connect-OMServer , Disconnect-OMServer
  • Alerts: Get-OMAlert, Get-OMAlertDefinition, Get-OMAlertSubType, Get-OMAlertType, Set-OMAlert
  • Recommendations: Get-OMRecommendation
  • Resources: Get-OMResource
  • Statistics: Get-OMStat, Get-OMStatKey
  • User Management: Get-OMUser

Update Manager

PowerCLI for Update Manager has always been available as a separate downloadable installer but it was hard to work with, you needed to work out which version of Update Manager you had installed on the server and install the exact version of PowerCLI cmdlets to work with it, this was obviously a pain when using multiple versions of vCenter or when trying to manage them from one machine.  In this version PowerCLI for vSphere Update Manager is no longer a separate downloadable component or installer and is now included in the core PowerCLI installer, it is selected by default during the PowerCLI install wizard which allows for simpler and quicker deployment and management of VMware products through PowerCLI. Enhancements have also been made to ensure a better backwards compatibility experience of this module and now supports versions of vSphere Update Manager all they back to 5.5.

vCloud Air

In the previous release we introduced PowerCLI for vCloud Air, allowing you to connect to your dedicated resources you have purchased and transfer your automation skillset into the cloud.  In this release you can now also connect to and manage vCloud Air On-Demand (vCA) instances with the –VCA parameter added to the Connect-PIServer and the new Get-PIComputeInstance cmdlet to list all available compute instances. Existing cmdlets for managing vCloud Director and vCloud Air can be used to work with vCloud Air On-Demand where applicable. Additional to these enhancements a new cmdlet has been added to make it easier and enhance the ability to work with OrgVDC Networks called Get-OrgVdcNetwork.

ESXi Host Hardware

This feature was asked for in the PowerCLI Communities, this is just one of the places we monitor to work out what our future releases look like.  New cmdlets have been added to work with ESXi host hardware, these cmdlets give the ability to interrogate your ESXi hosts and provide core system and hardware information.

Two new cmdlets have been added to work with this area:

  • Get-VMHostHardware lists host information
  • Get-VMHostPCIDevice lists all PCI Devices

Storage

Over the last number of releases we keep making enhancements to the storage cmdlets, with VSAN, IO Filters and now a number of enhancements have been made to the storage module introducing new cmdlets for working with VMware vSphere® API for Storage Awareness (VASA), NFS 4.1 and updated vSphere API for IO Filtering (VAIO) cmdlets.

The following new additional cmdlets are now available:

  • New VASA Cmdlets : New-VasaProvider, Remove-VasaProvider, Get-VasaProvider and Get-VasaStorageArray
  • New NFS Cmdlets: New-NfsUser, Remove-NfsUser, Get-NfsUser and Set-NfsUser
  • Added VAIO filters Cmdlet: Set-VAIOFilter

A pet peeve of mine and others

In previous versions of PowerCLI when you started the interactive shell it would always start PowerCLI in the directory where we installed the components, as I watched people use PowerCLI I would see them run a CD\ to get back to the root of the C: or they would use it from the install directory and often their commands would line wrap making it slightly harder to read, a small but hugely beneficial change was made in this release, the PowerCLI starting directory has now been changed from the full install path of PowerCLI to the root of the installation drive.

Business as usual

As usual updates have been made for PowerCLI to support the latest versions of the software it works with, in this case Site Recovery Manager (SRM) was updated to 6.1 and additional functionality has been added to discover SRM servers when they are deployed in a “shared recovery model” and support has been added for the vCloud Director 8.0 features which are provided by the backwards compatibility agreement of the VCD API.  All this and many more bug fixes and speed enhancements.

What an awesome release!