1812

I'm on 14,2 When you're changing core assignments, are you also changing vcpu static placement to match the number of cores assigned?

remove your topology and you can use anything you want up to 64 cores.

https://h20195.www2.hpe.com/v2/GetPDF.aspx/c04284193.pdf I would google around and see what others have put in as max procs as well. This includes looking at ebay ads too for complete working servers of the same model.

you can try following this (some you have done) as gridrunner talks about breaking up the bridge

Very perplexing. I was just reading about some issues with sata disk recognition on this gen/model and some were resolved by a firmware update... have you looked into that? I'll keep digging, but it's very weird. Once the "damage is done" can you switch back on the onboard raid controller and see if it can see all the disks in the array creation menu? (obviously don't make one, just see if they show up.)

fix for this log spam is on the first page.

Ok. Replicate the problem again making the drives disappear. Then immediately after they are gone, grab the diagnostics zip(tools>diagnostics) and post that complete zip file. That might help us know what is going on, and maybe why.

what hba are you using and is the onboard raid controller still enabled? Or are you using onboard raid, and if so, what model?

If your RMRR issue is the sound portion of a GPU, there is a workaround available in my sig.

My video editing server uses the onboard raid controller to manage 3 different higher performing raid groups without issue. But as stated, if something fails there is no notification. In 6.2.4 using the onboard controller and an hba caused the machine to hang or not boot. I have not retested in subsequent versions (didn’t have the need). there has been (for about a year now) in the feature request section for unRaid to have multiple cache pools. Maybe someday we will get it...

was the sapphire plug and play?

That's not 100% true. If you remove the topology from the xml, you can set any number between 1-64. I'm typing this on a 3 core Mac vm right now.

unsafe interrupts is now working, which allows us to unmask your real problem, that being: vfio-pci 0000:09:00.0: Device is ineligible for IOMMU domain attach due to platform RMRR requirement. Contact your platform vendor. This thread has some information regarding it. Some in the first post which may not be applicable on a G7 (using older bios vs updated-works on G6 servers...) On the second page, another user posted their steps using a custom kernel they made to bypass the RMRR issue with a patch via proxmox users. Currently these are the only 2 options that I know of to work around the issue. I have not attempted to use the custom compiled kernel.

it's not booting allowing unsafe interrupts (per the log.) delete all that you have in the syslinux and replace with the following: default menu.c32 menu title Lime Technology, Inc. prompt 0 timeout 50 label unRAID OS menu default kernel /bzimage append vfio_iommu_type1.allow_unsafe_interrupts=1 initrd=/bzroot label unRAID OS GUI Mode kernel /bzimage append pcie_acs_override=downstream initrd=/bzroot,/bzroot-gui label unRAID OS Safe Mode (no plugins, no GUI) kernel /bzimage append pcie_acs_override=downstream initrd=/bzroot unraidsafemode label unRAID OS GUI Safe Mode (no plugins) kernel /bzimage append pcie_acs_override=downstream initrd=/bzroot,/bzroot-gui unraidsafemode label Memtest86+ kernel /memtest once done verify the sound card is in it's iommu group by itself and try again with the vm. if it fails, post your new diagnostics immediately after.

post your diagnostics zip file (tools>diagnostics)

i'm on 6.5 rc3 and run 2 osx vm's no problem using pc-q35-2.11

the xml you posted shows topology defined and shows 8 sockets with 2 threads.... you can try adding an emulator pin on cpu 0, but with 16 cpu's, I don't think it'll make that much of a difference. What else is running off the NVME drive? downloaders/dockers? Something is bogging it down and I don't believe it is cpu related since you're throwing 16 threads at it (at the moment of typing, I'm running 2 osx vm's on a W3520, with each vm on it's own ssd and only 4-6 gb of ram, using gtx 710's, so way less power but runs smooth)

My server rack with that particular setup is currently torn down and packed for moving. It will be about 2 months before I'll have it back up and running. I can try to do this at that point, if I can remember!

You need to start with VM passthrough problems on the previous page dealing with unsafe interrupts. Without allowing unsafe interrupts (append vfio_iommu_type1.allow_unsafe_interrupts=1 initrd=/bzroot) , you will not get any closer to your goal of a working vm.

pass through a GPU and watch it crawl when adding large amounts of ram.

This is the best place to do it, so that if anyone else has the same problem on a proliant, then they can find the answer here. Let us know the error and upload your full diagnostics zip as well. (tools>diagnostics)

and after that option is implemented, some user will accidentally disable it, copy over something important to a single cache disk where it will sit for a week, and then the disk will die losing their data and they will blame unRaid for their error... ¯\_(ツ)_/¯ Much easier to just create a cache only share. If you created a folder that exists on the cache only without making it a user share, then you probably took the same if not more steps doing it that way to create it.

I can confirm this works, though it might take a minute or two for the names to appear.

First off, this looks awesome. I set it up as shown in the video (I believe), and the pi-hole dashboard shows it is routing dns quarries from my pfsense vm to pi-hole, but not blocking anything. I even added the same single block list like you did. But when I go to the same Forbes site, it still shows ads. Any ideas what I did wrong? ---- edit: Figured it out- needed to force DHCP renew on clients to get the new DNS server entry. After that, works like a charm!