For the dynamic memory alloction to work on a specific VM you need to have installed on that VM the guest tools / agent instlled? Otherwise the dynamic memory will not work on that VM and the system will allocate the maximum static memory to it and lock it. (Like a thick provision)
Ok, ok, I’ll try XCP-NG. Hope it works well on our Dell R740xd. Love that there is now a free XO, can’t wait to take it all for a spin against my daily driver vsphere 7.
Thanks, interesting. Do you feel that the various guest drivers for Windows handle the dynamic RAM differently? I'm not sure but I've had som issues that have led me to try XCP-ng vs Citrix vs the 9.x from Xen.
You've covered just about everything with XCP-NG and XOA, except GPU/PCI passthrough. You should really show a video on this! I've done GPU passthrough with limited success. The one time It was successful. Windows would work via RDP, but could never get a Signal out of the GPU ports unfortunately.
@@LAWRENCESYSTEMS Tom: Your series of videos on XCP-NG is actually quite complete. I can imagine a scenario where people might want to, for example, set up a Plex Media Server at home, on their home lab, and might want to passthrough a GPU to help with GPU accelerated transcoding, as an example. I think that the way that you tend to cover these topics -- presents it in a way that is easy and simple for people to understand, so I can certainly see how that can be valuable to the community-at-large.
Is htop ballooning aware or is the so really seeing ram resize? I always thought dynamic is only using vmm ballooning. Is that a xen driver in the guest os doing that? BTW i think other Hypervisor can also „hotplug“ RAM so static would be changeable at run. BTW It’s ironically that sizing down slows it down, unneeded in this specific config
@@LAWRENCESYSTEMS I see. So they "barely saved from removal"? Their statement is below for anyone reading. DMC (Dynamic Memory control): in fact, this first feature isn't removed. Let's say "barely saved from removal", because XenServer simply nuked it. We decided to keep it enabled because we know that there are users who depend on it, despite known issues. However, we advise not to use it unless you can't really do without it. And don't set the dynamic min limit too low, unless you want VMs to crash when the hypervisor will tell them to give back RAM that their actually need. We might have some R&D work to rewrite it correctly in the future.
re: overprovisioning for CPUs I always assumed that you can overprovision your CPUs so long as they're not all trying to use all of the CPU cores that you've given to the VMs, at the same time. I've also always assumed that if your VMs start using more of the CPUs each, such that the total number of CPUs used is greater than either the physical or the logical number of cores, that you would start to run into performance issues and/or eventually that either tasks/process/VMs would start to crash. Thank you.
What a great thumbnail. Love it.
Great video Tom! Never fully understood how this process worked; would love to see the processor video as well.
Looking forward for the CPU cores over allocation!
Thank you for this video. I've been meaning to look more into this myself, but hadn't got around to that yet.
Hello Tom. Great video, supper informative!
I sure know that I'd like to see the same video for processors!
Have a great day
For the dynamic memory alloction to work on a specific VM you need to have installed on that VM the guest tools / agent instlled? Otherwise the dynamic memory will not work on that VM and the system will allocate the maximum static memory to it and lock it. (Like a thick provision)
Ok, ok, I’ll try XCP-NG. Hope it works well on our Dell R740xd. Love that there is now a free XO, can’t wait to take it all for a spin against my daily driver vsphere 7.
Processor overprovisioning please. I'm getting closer to deciding to switch to xo from libvirt the more I see your demonstrations.
Thanks, interesting. Do you feel that the various guest drivers for Windows handle the dynamic RAM differently? I'm not sure but I've had som issues that have led me to try XCP-ng vs Citrix vs the 9.x from Xen.
Ballooning driver for Windows works similarly but you should not forget to install compatible vm tools.
@@berndeckenfels Which tools are these?
You've covered just about everything with XCP-NG and XOA, except GPU/PCI passthrough. You should really show a video on this!
I've done GPU passthrough with limited success. The one time It was successful. Windows would work via RDP, but could never get a Signal out of the GPU ports unfortunately.
It's just not something we ever use which is why I have not covered it. It's pretty low priority for me
@@LAWRENCESYSTEMS
Tom:
Your series of videos on XCP-NG is actually quite complete.
I can imagine a scenario where people might want to, for example, set up a Plex Media Server at home, on their home lab, and might want to passthrough a GPU to help with GPU accelerated transcoding, as an example.
I think that the way that you tend to cover these topics -- presents it in a way that is easy and simple for people to understand, so I can certainly see how that can be valuable to the community-at-large.
Is htop ballooning aware or is the so really seeing ram resize? I always thought dynamic is only using vmm ballooning. Is that a xen driver in the guest os doing that? BTW i think other Hypervisor can also „hotplug“ RAM so static would be changeable at run.
BTW It’s ironically that sizing down slows it down, unneeded in this specific config
HTOP is just reporting the memory available, the xe-guest utilities in the guest OS are working with the XEN hypervisor to make the changes in memory.
This is unfortunate timing. Dynamic memory control is being abandoned or planned for removal as of version 8.3 alpha.
No, it's not. Read their blog post
@@LAWRENCESYSTEMS I see. So they "barely saved from removal"? Their statement is below for anyone reading.
DMC (Dynamic Memory control): in fact, this first feature isn't removed. Let's say "barely saved from removal", because XenServer simply nuked it. We decided to keep it enabled because we know that there are users who depend on it, despite known issues. However, we advise not to use it unless you can't really do without it. And don't set the dynamic min limit too low, unless you want VMs to crash when the hypervisor will tell them to give back RAM that their actually need. We might have some R&D work to rewrite it correctly in the future.
re: overprovisioning for CPUs
I always assumed that you can overprovision your CPUs so long as they're not all trying to use all of the CPU cores that you've given to the VMs, at the same time.
I've also always assumed that if your VMs start using more of the CPUs each, such that the total number of CPUs used is greater than either the physical or the logical number of cores, that you would start to run into performance issues and/or eventually that either tasks/process/VMs would start to crash.
Thank you.