Is it possible to create a new VM from a physical machine, the way Parallels Desktop and hyperV can? I have an ESXi environment provided by outsourced IT who really don't know what they are doing, and I am trying to insource as much as I can. One thing these IT "professionals" do is bill 10 hours to provision new machines manually, including all the software install. They want to play the same game creating VMs, so I am hoping to find a way to simply image one of the new physical machines they provided, and use that as the gold image for some software install testing. Eventually having a VM based on a PM that has been through the ringer would also be helpful, but not on the joke, 10 year old poorly configured hardware we have now.
There is a P2V convertor available - works OK in the most part but beware of specific drivers etc based on the original machine, in essence you will be creating a VM based on the PM which carries over with it some 'skid-marks' of the PM that the VM doesn't need or can be problematic.
Think of it as a quick fix short-term solution rather than a long-term production machine.
Also, if you are paying for creating of 'golden images' worth bearing in mind that these will need patching and updating otherwise everytime you deploy from these templates you will be faced with the updates/patching required at the time you deploy.
I tend to have an OOBE template of most OS versions then a customer-specific version (AD membership, default AV, patching software install etc) that I revisit at least once a year to apply MS or other OS specific security patches etc to update accordingly. Just saves time when deploying new VMs.
Clean is always better than converted but when time is an issue conversions can work but beware of leaving them to get worse.
So that's a P2V process.
VMware's tool (Converter Standalone) has been discontinued but a new version is in development.
Golden Image may actually not be a good description of what I am doing. It's not for spawning multiple production VMs. It's purely for testing an automated software install procedure, which will then be applied to a hundred plus physical machines.
One limiting factor is VMWare's approach too Snapshots. It seems like they really are for very specialized use, and certainly not to be maintained long term. I do my own testing in Parallels Desktop in an iMac Pro, and I have a VM right now with s year old snapshot of the base OS, and then 20 different child snapshots, one for each 2023 Autodesk product, and all those snapshots then have their own child after launching the software as a standard user. And restart times are less than 30 seconds, create new snapshot about the same, and delete snapshot less than 1 minute. Imagine trying to do that in VMWare, even on a good host. Now imagine doing it on a 10 year old hand me down server with out of date drivers and a config planned by a drunken ape. It's... painful.
Which is why I am looking into cloning a physical machine. We have been asking for fit for purpose host hardware for three YEARS, and this useless IT company has done nothing. So we have sourced an older BIM/CAD machine, and I am trying to convince the customer that the IT company doesn't know anything, and we can install and configure ESXi in less than an hour, and have a fit for purpose test bench. But then we either wait three weeks for IT to create a new VM, or we clone an existing physical machine. And thankfully we have a few brand new physical machines that we could clone, so it would be a "clean" VM. I would rather have IT build it, but I just found out today that their "professional IT clown" approach is to use PXE with a 5 year old Windows 10 1703 ISO, then MANUALLY do the windows up[dates, including updating to current Win10 build, then install all the office standard software after renaming the computer and joining the domain. They have managed to turn what should take a person about 10 minutes to kick off, then go do useful work while the automation does it's magic, into 10 hours of billable time. That it takes weeks for them to schedule. That doesn't get done the same way every time because how could it. If it was me, instead of paying these idiots I would be be showing them the door with extreme prejudice.
In the meantime, since I have remote access to one of those machines, I just tried to download VMWare Converter. I can search on the VMWare site, and I get a link for VMWare vCenter Converter Standalone 6.2.0, but clicking the link just takes me back to the list of all the software, where I did the search in the first place. Is this just VMWares way of saying "no longer available"?
Ah - my bad, according to this https://blogs.vmware.com/vsphere/2022/02/vcenter-converter-unavailable-for-download.html
It's not longer available - doesn't mean to say it won't work though so here is a link to a download - just pulling the installer down from an old source I've used before - once it's down I'll post the link.
Also - you don't seem to be getting Snapshots usage right - if you'll pardon the bluntness
These are designed for short-term usage only - many reasons, storage consumption only one but tomb-stone-ing more relevant in an AD environment.
Would suggest that you take clones rather than snaps of your various software variant installs and bump these off to less expensive storage where possible (e.g. non-live external HDD etc) this will give you the ability to work on the 'latest good' or revert to the 'latest good' in the event that something or some patch causes a problem.
I have felt your pain working with legacy stuff, there's always a way, even if $ is a limiting factor, you just need to get imaginative, virtualisation gives you options.
Link will follow to the P2V converter that I have used before - it's a VMware installer but feel free to check it before you install - I would!
You rock! Thanks. I downloaded the Starwind Converter too. One or the other will work, though at least Starwind requires writing direct to the ESXi server, so I'll wait to try that.
Oh, I see now that it's not a good workflow in VMWare. Or at least ESXi. I wonder, does VMware Workstation have the same issues? Certainly it's a stellar workflow in Parallels. Works like a champ. I have had upwards of 60 snapshots of a single VM, for software data validation purposes, and it worked a treat. Moving back and forth between different saved snapshots is trivial, and fast. Works well in Hyper-V too. But obviously both of this packages have a very different concept of Snapshots. I'll take a look at clones in ESXi, especially since even just one snapshot seems to be enough to make a VM mess the bed on this 1 hamster on quaaludes Host we have now. I would have expected the IT clowns to explain that snapshots on ESXi aren't viable and use clones, but they wouldn't know a good workflow if it hit them upside the head. 🙂
Try this link https://1drv.ms/u/s!Ar8T-neH3QIWtHlpTzCV_um0HYeh?e=Nv6IJP - should be good - as I said, straight from VMware back in the day but feel free to scan it before you use
OK so snapshots v clones ...
delta changes are stored in snapshots so the longer you leave then around the more space they take up
This causes issues eventually 2fold - one, space consumed, note IMPORTANT, that this space isn't visible, alerted on via v-Sphere so can cause issues at the storage layer, two tombstone isn't just a Hollywood scary thing, if you rely on snapshots for over 90 days or so to revert back to then you can find that they aren't as useful as you think, the VM will simply not just re-join to the AD.
Anyhoo, try running the VMconverter on your PM as it stands (local admin privs etc and vCenter creds etc) you should find that it will create a VM of your PM - it may time-out if you have a slow network between PM and virtual infrastructure dependant on size but if you pick your low I/O time well it should succeed, the resultant VM will be probably messy but should run and will be a direct copy of your original PM.
Thanks for the detail! Good to get a better understanding of what's going on. I will definitely be minimizing my snapshot use going forward. Unfortunately we have a naked vShere environment, no vCenter Server, so creating clones is less than convenient. But if it gets me through this effort, I'm ok with that. I suggested to the customer that they just install Hypoer-V on a workstation from the getgo, but they listened to the IT clowns pushing ESXi. Now I think because they wanted the billable hours making VMs. They could have saved SO much money just buying a retail Windows 10 Pro and making a VM on Hyper-V. Oh well, they could have saved a ton firing this company and getting someone in who actually supports the firms work, rather than bleeding them dry while doing the absolute minimum.
I'm going to have to look into why Parallels is so different. there has to be some kind of tradeoff to have snapshots behave so differently, and at least for my use case, better. I am curious, is VMWare for the desktop the same? Or being a desktop hypervisor is the use case more similar to Parallels and thus whatever benefits the ESXi implementation brings doesn't really apply?
Creating Linked Clones on an ESXi standalone host is super easy.
But to disapoint you - you cant get linked clones without snapshots !!!
Back in times of esxi 3.5 I created a howto - all you need is to get over your aversion to snapshots, install WinSCP on your admin host and 20 minutes later you should have the first linked clones up and running.
All that has changed since then is that you nowadays would use the embedded editor of winscp for all edits. Do not download files to your adminhost and edit them there !
Before you ask:
no - you cant update the original basedisk to rollout a Windows update to all your clones- if you try that you damage all the linked clones.
I'll have to look into linked clones, thanks for the link! Weekend reading. 🙂 As for windows updates pushed to all clones, that's the LAST thing I would do. The only thing more likely to make an Autodesk product unstable than a windows update is an Autodesk update. The whole point of snapshots for me is being sure to get back to a VM state before either MS or ADSK broke things, usually so I can then find out what and how and find a workaround, since neither MS or ADSK seem to care. On Parallels I install windows updates once a month after creating a Last Known Good snapshot, and then delete the three months old LKG. So I always have two snapshots to fall back to if something suddenly goes south. And I can quickly create an AFU snapshot too, so I can quickly (seconds) jump between good and bad conditions and research the issue. It's glorious.
Make sure that you strictly separate the disks of the linked clones.
Use one disk (rather one or more snapshots) for the Windows and AutoCad installation.
Use a completely independant disk for all the data produced by the users (autocad templates, drawings, user-data ...)
This way you will never lose the real important stuff when something goes wrong with the installation.
And important: educate your users to ignore all "consolidate me cries" from ESXi.
"The only thing more likely to make an Autodesk product unstable than a windows update is an Autodesk update"
LOL - what happened to software .... and how come all vendors still tell us to update to the latest and greatest as a first step in troubleshooting ???
By the way - if you need help with setting up linked clones let me know via skype.
Good to know about the consolidation. I have been tempted by that message lately as I search in vain for a way to turn this ESXi turd into a way to get some productive work done. But I should say again, no users will ever touch the VM. It exists purely for me to test installs of software, with the ability to roll back to exactly the conditions before the install. Once the process is refined, it is applied to a firm full of physical machines.
Discussion about linked clones goes on here:
I finally had a chance to try this, and I ran into a problem where the utility is trying to use SMB1, while Windows 10 has now deprecated that. Do you know off hand if that is something I can configure in the utility, or is it a limitation of this older version? In which case I guess I hope someone else might have the latest version laying around and be willing to share it.
Also, a question on procedure. I have access to a Hyper-V P2V utility, and I tried that as well, and because the physical machine I have is a single drive machine, I was writing direct to the network. That failed pretty dramatically. But that begs the question, is this really a bad workflow, and the better option is for the customer to by an external USB3 drive, and we clone to that, then deal with getting the VHD on the host? That feels like the right answer, especially since the network provided by the IT clowns is pretty bad. While copying a VHD between hosts with SCP the network performance metric was usually in the 60-80MBS range, on a gigabit network, but at times it dropped to only 5MBS. I am unclear at this point if that is normal on a well managed network, or a(nother) sign that the IT clowns aren't doing a good job.