meep

My go file is likely around 10 years old, so there may be several flukes holding things together!

Thanks. After a frustrating afternoon of troubleshooting and restarts, the root cause has become annoyingly clear. It seems that the update from 6.10.3 to 6.11.5 caused management access settings to revert defaults, so I lost my custom port and ssh log in ability. Grrrr I'm sure this fact is buried in release notes somewhere and there's good reason for it, but it caught me out big time. It's the first time since the 4.x days that an update has gone somewhat awry, and that includes three complete system hardware swap outs, so all in all, not a bad run. I'll try the update again later in the week and see how it goes.

Happily running on 6.10.3 for many months and finally updated to 6.11.5 via WebUI. On reboot, I had no network access to my webUI (192.168.1.199:1234). I attached a display and saw the system was booting as far as the point where my GPU is handed off to the system for VM use. I tried booting in safe mode, but the boot process stalled at an AMD VI error (not noted). I could not ssh into the system in either case. I took the USB to another machine and restored to 6.11.1 from the 'previous' folder. Now, my system boots and I know it's operational as I see a VM on the GPU and my Kodi clients can access media through Emby Server docker on the system. However, I have the following odd behaviour (all on 6.11.1 after restore from 6.11.5); I can successfully SSH into the server as root during the boot process. SSH user seems to be booted out somewhere during the OS startup (so I have ssh access for a very limited time) After startup, no SSH is possible (connection refused) Web GUI is not accessible at any point I can ping the server at any time, and it appears to have started up correctly as VMs and Dockers are operational, but docker UIs are not accessible I can sometimes access shares to mount from machines on the network, but such access needs to be instigated afresh - any previous mounted shares fail and I must (on MacOS), select connect server and enter SMB URL. Even then, connection is very slow to initiate and access is glacial. I've tried, without success; deleting 0kb SSH keys restarting my network router (Unifi USG) restrating unRaid server I did manage to manually edit VFIO to remove GPU binding and generated diagnostics in safe GUI mode. I guess if I can get this figurerd out for 6.10.3, I can apply learnings to 6.11.x tower-diagnostics-20221203-0535.zip

I had a big roundy birthday recently and it started me thinking about all the consumer technology I've seen either invented or popularised so far in my lifetime. Highlights include; VCR Mobile Phone Compact Disk Home Computers Unraid Internet Electric & Self-Driving vehicles Solar PV I'm sure there's a ton of stuff I'm forgetting, including advancements in medicine, science etc. I love telling the kids about the times before all of these. But I also am excited for the future, or as much of it as I will see. What are the big ticket items we can expect or anticipate in the next 50 years? (assuming we don't destroy the planet before then).

Just picked up a Ryzen 5 5600G on Amazon.co.uk for £129.98, down from £289.50, a 55% saving. They also have other bargains like Ryzen 5 5600X for £182.99 (-35%) and Ryzen 7 5700G for £204.99 (-49%). Lots of other variants as well. Good deals if you're thinking of building out an unraid server.

Are you sure that you are mounting the unRaid share in Debian, and not one of the share folders on one of the array drives? How many drives are assigned to your share? The above behaviour could happen if you mounted a share folder, instead of the share. When unRaid creates files, they could be written to a different disk and therefore would not show up in Debian? Can't think of anything else.

Hi @Ricardo Maranhao I managed to cobble together a quick tutorial. Let me know if that's clear, helpful, or if you have any further questions that I could address in the article. https://mediaserver8.blogspot.com/2022/05/mapping-additional-and-network-drives.html

Not off hand. If I get a chance I'll try too write up a short blog post later

Yes, inside your vm, you can mount any unRAID share just like you would from a regular PC. It behaves just like a network volume. This is a good solution. I run blueiris in a Windows VM for my security cameras. I have an unassigned disk passed through to the VM for recording storage / scratch The VM sees this as a locally mounted HD. I also have a an unRAID share mounted in the VM as a network drive. Here, BlueIris moves ‘motion’ tagged clips nightly. Works really well.

Qcow2 images don’t auto-shrink when data is removed, afaik. i found this on the inter web describing how trimming is possible if the vm is set up just right. I’ve never tried it in unRAID, but it should work. There may be other solutions as well; https://kparal.wordpress.com/2017/10/06/automatically-shrink-your-vm-disk-images-when-you-delete-files/

Nvme drives run hotter than other drive types. This is normal. If you are worried, you can add a heat sink to the nvme drive. in unRAID, you can adjust the warning temps for individual drives on the disk settings tab. Look at the SMART section. I have warning and critical temps for nvmes set to 55 and 65 Celsius respectively. This way, you prevent worrisome but unnecessary warnings in the UI.

the bus/slot assignments within each vm are independent of other VMs and relate to virtual hardware of that specific vm only. You seem to be saying that you are striving to use unique slots across different vms? This is not necessary. You can use bus 0, slot 5 in all vms if you want. You only need to take care that you do not try to pass the same physical device in your server to more than one vm that you intend to run simultaneously.

Folks, been using MacInABox (clover flavour) for a long time with no problems. However, recently the SSD I was using for the VM died (complete failure). Needed to re-install and retrieve files from TimeMachine, I updated MacInABox and successfully got my Catalina back up and running on OpenCore. (AMD RX580 GPU passed through). Except it's not stable. The main problems are; 1. Network access to unRaid shares is glacial, though internet browsing is OK. (I have e1000-82545em configured, no issues accessing shares from a real Mac) 2. Programs (in particular Adobe Illustrator) is causing the OS to freeze regularly 3. Playing any video file in a browser will always cause the OS to freeze after 15-20 seconds. Rebooting the VM fixes the OS freeze, until the next time. I'm not at all familiar with OpenCore so am really stuck as to where I can even start to troubleshoot. Rock Solid on Clover. Multiple Problems on Open Core. Any insights? Back to clover?

Yep, this is the time to buy. 18TB showing at €309 for me in UK store. Bought a few 14TBs last year and they're great.

Hi @jbartlett Suggested Feature: So one thing I struggle with in my server is keeping track of what the various drives are used for. unRaid itself does not provide a way to annotate drives or add a memo etc. In my case, I have a number of unassigned SSDs and other drives passed through to various VMs and for other uses. Diskspeed docker does a great job of visually representing the drives, and the capacity overlay is super useful. For me, a great addition would be the ability to manually add an additional text memo to some / all drives to help me quickly remember what that specific drive is used for (without needing to sift through VM configs etc.) I've visualised something here to get the idea across; You can see that I can determine from the titles that most of these are array drives, but those on Ports 6 & 7 are not. In a few months after setting stuff up, I'd love a handy reminder of what those drives are doing in my system. You see here that I mocked-up an additional text overlay where I can set a reminder for these specific disks. It wouldn't even need to be a separate field. If the existing disk size overlay could be amended to take custom text, that would work too (I'd be happy to manually add the capacity to my custom label) For me, this would be a great feature. Thanks for considering it.

The behaviour sounds like a situation where the GPUs have been reserved. In this scenario, the boot sequence shows up to a point, but appears to stall. Have you a clean unraid install, or have you ever set any VFIO reservations for passthrough? (if you have, and the n shifted or added devices, the assignments can change and th wrong devices can become reserved).

That should be fine. I believe the recommendations are to give unRaid at least core 0, and alway include core 0 as there is some benefit in that (i'm not sure what). Remember, the above guidance is specifically around isolating cores for high performance VMs such as gaming. If you have a few dokers that are not computationally intensive, you can pin non-isolated cores to them, leaving a bit of headroom for unRaid to take as needed. Also, you don't need to isolate ALL the cores assigned to a VM. You can, say, assign 4x threads to a VM, but only isolate 2 of them. The VM then has two dedicated threads form its exclusive use and two others that may or may not have other tasks assigned. There's also a TON of esoteric analysis on here about performance benefits of assigning specific threads / cores to VMs based on how the CPU is constructed etc. Its as bit over my head, but some peeps seem to have eeked out a bit more horsepower by paying close attention to how core assignment is managed. based on CPU architecture.

@brainbudt , @sbo365web If you switch on the help in the GUI on the isolation page, you will find this; So, when you pin a core to a VM or Docker, that core is allocated to the VM/Docker, but the unRAID OS may still access it and use it for various tasks. When you isolate a core, it is no longer accessible even by unRaid for routine tasks and should therefore be 100% dedicated to wherever it's pinned. You can pin cores without requiring a system restart. You must restart your unRaid server to effect any isolation changes.

So as someone who's just finishing up a trial, it looks like I missed all the pain. For me, it was a very straightforward initial setup ad the update was smoothness itself. The 1.8 UI, on first impressions, looks like a vast improvement on what went before which was somewhat clunky and tricky to navigate (on desktop). Now, my only decision is Qobuz or Tidal. (I'm leaning towards the former).

Hi @Jagadguru Yes. I use these with great success. See here; https://mediaserver8.blogspot.com/2020/04/unraid-discrete-usb-passthrough-to.html and here; https://mediaserver8.blogspot.com/2020/04/mediaserver-83-bifurcation-edition.html You can even assign cards in each slot to different VMs as they can show up in different IOMMU groups. They work using a technology known as PLX. Essentially, theres a traffic cop chip on them that assigns each slot on the expander board a slice of time on the motherboard slot. In effect, this is transparent and the system sees cards populated in the expander as discrete and additional devices. I've encountered no problems using up to 4x USB 3.0 cards in these expansion devices, all assigned to different VMs simultaneously.

Yes, a 4x -> 4 1x would be useful, however given the way the PLX chips work, it just slices the time each card gets in the slot so it would likely be a different hardware solution in that case. This is the one I'm currently using. It's really compact and I can mount it anywhere in my case, running I/O brackets to the back panel if necessary. That way, you only take up the number of precious expansion slots you actually need, rather than all 4 (or 8 in the case of the lorger expanders designed for dual width GPUs).

I have a windows vm running blue iris for Security camera duties. No video card. i also use this as a whole house audio server with multiple legacy m-audio delta multi-channel sound cards. No Linux drivers, but works fine on windows or Mac OS. No video card required. this also runs windows-only tv tuner software for whole house DVR, no video card required. i also have this connected to an old alarm system for which I wrote some drivers. The vm manages the connectivity and interfaces with a node red docker where all the logic is implemented. No video card required on desktop os for this job either. there are loads of use cases for desktop vms with no gpu. Of course, a lot can be handled in other ways on Linux, for example, but for those not comfortable in the black arts of bash and the command line, a headless windows or Mac vm can be a great tool.

I have a link to a blog article in my sig that has a few steps worth trying. You never know, something there might help in your case.

Great to hear you got it working. For future reference, as requested here's a short blog post from a while ago that has details of the USb audio devicves I use; https://mediaserver8.blogspot.com/2019/08/vm-audio-solution.html

Ah, that makes more sense. You don’t have audio in your streaming client, not the physical card. I’d imagine this is because there’s no physical external device attached to the decode the audio stream does this kind of configuration work with a bare-metal (non vm) set up? I’m not sure those dummy plugs are smart enough to fool the system on the audio as well as video side of things. i do sometimes use parsec, but also have cheap usb audio devices attached to VM and that seems to do the trick. (Usb device plugged in to USB adapter which is passed through to the vm).