sonofdbn

This morning I found my server (6.12.6) had hung: couldn't access the GUI or the shares. IPMI showed an unRAID screen with a field for user name, but I couldn't type anything into the field, even using the virtual keyboard. In the end I had to reboot, so there was the dreaded unclean shutdown, and now parity-check is running. Don't think I've done anything drastic, although over the last few weeks fiddling with docker container access I switched from ipvlan back to macvlan and vice versa a couple of times, but the server has been running fine (on macvlan without any further changes) for at least a week. Hoping someone can make sense of the diagnostics and tell me what caused the problem. tower-diagnostics-20240224-0945.zip

Did I mention the great support in the forums? Now let me find something else to complain about.

I like saving money as much as anyone else, but where I am, the current Pro licence is roughly the cost of an 8TB hard drive. If I want to avoid subscription/extension fees and go for a new Pro/Lifetime licence, I'm guessing it will cost less than a 16TB hard drive (I currently have four in one array, two in another). Of course unRAID isn't perfect (I absolutely hate the clickfest Alerts/Warnings/Notices system in the GUI), not all my queries here get answered and I will never understand the difference between macvlans and ipvlans, but me and unRAID have come a long way since I wrestled with PATA drives and thought a 400GB drive was huge. I was going to say I'm not a fanboi, but actually I think I am 😀

Specifically re Firefox on Android on a tablet: when I use FF there is no tab bar, which I find makes it almost unusable. I did try an extension, but it could display tabs only on the bottom. Am I missing a setting? I'd love to be able to use FF on the tablet so that it can sync with my PC. I've seen occasional advice about not using FF with the unRAID GUI. Here's one I managed to dig up: In practice I've been using FF almost all the time with no problems, but have just kept in mind that if there's an unRAID GUI related issue it might be a FF issue.

Today Firefox (latest version) decided to show black screens. Only the FF windows; Vivaldi is fine. Turns out I had to turn off hardware acceleration, so that fixes it, but why? What's changed? Certainly not my hardware. And I haven't updated FF since last night when it was working fine. But yes, the password savers/managers are great. If only FF didn't insist on positioning password options right over the dropdown box where I need to select something on the airline website. It's really ridiculous that I have to choose a browser depending on which website I want to go to. I exaggerate, of course, but only a little.

Did you ever look at Synology Photos? A long time ago I used the old DS Photo (?) from Synology. It wasn't a good experience, but at first glance Synology Photos looks quite nice. While I do think that Synology hardware is over-priced, if you can make good use of their software the price is much more palatable. unRAID is much more flexible of course, but as this thread shows, we could do with better apps in some areas.

@rheumatoid-programme6086 A bit late to this, but just wondering exactly what you did to run unRAID after you switched from the USB HDD DAS enclosure. You mention the 6x SATA adapter in the NVME slot, but since you're running a mini PC, where do the hard disks go? (I'm assuming the mini PC is one of those small ones that don't have any HDD slots, like this HP Elite Mini 600 G9 Desktop PC.)

I'm on 6.12.6 and am trying out virtual-dsm. It seems a little different from other docker containers: it's essentially DSM running in a VM, installed as a container. I've got it set up using custom network br0, which is a macvlan, but I don't know how to get Swag/Letsencrypt set up to allow me to reverse proxy to it. For example, if the container sits at 192.168.1.100, how do I get mydomain.duckdns.org to point to that IP address? Swag doesn't have a template for virtual-dsm, so I've tried various templates, but no joy.

I tried out btrfs on a cache pool some years back - maybe 5 years? Not sure if cache pool is the right term - basically I had two duplicate SSDs in a RAID-1 type arrangement - this was before unRAID allowed multiple cache pools, I think. I followed the official setup process, so no weird customisation involved. I ran into some problems with the filesystem and spent some time trying to fix it with help from this forum. I can't remember whether I was actually able to fix things, but whatever the outcome I do recall that my impression at the time was that it was just too risky for me to continue with btrfs. So I went back to a single cache drive on xfs and have never used btrfs since then. In fact I might have done what you did and reformatted the drives as xfs. (Unfortunately I can't find the thread now - but it's long outdated.) I do believe that btrfs has matured since then and should be better. And it does seem that lots of people use it quite happily. My understanding is that at one time if you wanted redundancy/RAID 1 type protection on any pool device (in effect duplicating the pool drive), you had to use btrfs; xfs didn't support this. Perhaps these days you can do it with zfs, but that's too complicated for me. For me, one important and sometimes overlooked feature of any data protection policy is that the user must have a high degree of confidence in the system. Whether this confidence is subjective, justified, etc. is in some ways immaterial. You want to be able to rest easy. I just couldn't trust btrfs after my experience with it, but I do accept that it's perfectly fine for others.

I'm finding it surprisingly difficult to settle on one browser to handle my needs, which I don't think are overly demanding. One thing I really want is vertical tabs - with current monitor aspect ratios I would much rather have my tabs using up the width rather than the height. Admittedly some of my other prejudices are largely subjective, but here's a brief summary: Chrome Don't entirely trust Google (and yes, my phone is an Android phone), but it's more the sneaky things they seem to try with privacy settings. The one thing where Chrome is really good is its auto-translate. Other browsers do it, but not so seamlessly. Edge Can't abide the constant pushing of Edge and Microsoft services. Firefox My current main browser. Quite nice, but doesn't play nicely with unRAID sometimes (maybe this has changed?). Also, although it doesn't offer vertical tabs natively, the Tree Style Tab extension is good. It has the unfortunate quirk of occasionally opening a tab in a new window when I click on the tab - don't know if it's the extension or Firefox or my mouse-clicking ability - so I end up having to drag the tab back to the original window. Not a big deal, but happens often enough to be mildly annoying. Firefox also has this truly frustrating habit of popping up its password suggestions right over the part where you need to see or select an item. I travel reasonably frequently and when I try to pay for my air ticket it throws up password suggestions right on top of the dropdown box where I have to choose a credit card, so I can't see any of the credit cards listed. Vivaldi Previously my go-to browser, largely because it has native vertical tabs. (It might be the only browser with this feature.) But over the past year or so there have been additions that seem to get in the way (like Workspaces), and I get odd behaviour. Right now, it won't launch - probably need to reboot my PC. When it was the default browser it wouldn't open links, and on some websites it won't download files. Sometimes tabs go missing in the tab bar, although they're still active. On the Vivaldi forums I always read that it must be because of extensions, but I don't think I have anything too weird: Ublock Origin, Bitwarden, Google Translate mainly. Vivaldi also has this annoying habit of resetting the address bar spacing and the search engine when it updates (no way I would have selected Bing as the default search engine.) Others - Opera: used it way back when it was the new kid on the block, but when I looked at it recently it seemed to have some odd marketing gumph and it didn't seem to be worth trying to work through it. - Safari: does it still exist for Windows? - DuckDuckGo: nice idea in terms of privacy, but I found it klunky and didn't pursue further. - Bravo: never tried So currently I use Firefox on my main PC, with Vivaldi as a fallback (and for booking plane flights and unRAID). Annoyingly I also have to use Edge because some corporate thing only seems to work with Edge. I also end up having to use Vivaldi on my tablet because Firefox Android doesn't think people need tabs. So syncing my browsing is essentially impossible. I end up sending links to myself or via Joplin. I sort of miss the (perhaps not so) good old days when it was either Internet Explorer or Netscape Navigator. I don't test out browsers frequently, so some of my points might be out of date. And I'm sure there must be ways of dealing with the various irritations, but it's rarely intuitive or easy. (For example, try adjusting the spacing on Vivaldi's address bar.) All of this makes me think that I might surrender to our Google overlords and just use Chrome on everything. Resistance is proving to be quite futile at the moment.

I managed to install a container (virtual-dsm) and it seems to be running fine. Now I'd like to set up a reverse proxy to the container via Swag. There's no prebuilt nginx config file, but I'm happy to fiddle around with this. Is there anything different about a docker container installed in the "normal" unRAID way (download from Community Apps and click install) vs a container installed via docker compose manager? Particularly anything that might affect what needs to be done for a reverse proxy? (I'm not well-versed in linux, docker containers or reverse proxying.)

@Cirion Would be interested to hear your experiences with Immichh if you go ahead. I tried out the docker a few months ago, and it seemed quite good, but it missed a few photos on importing. Back then and possibly still now, the author of the software warned that it was under heavy development, and I found that after an update some things were broken and some significant reinstallation was needed. Since I need something that was stable and reliable for the family to use, I didn't continue with Immichh. It does seem like a great solution, but I'm going to wait until it's a bit more settled.

I do have SMART warning on 6 out of the other 9 disks, but they are all UDMA CRC Error count warnings. I suppose it sounds bad, but from what I've read, this type of error is usually an indication of a loose wiring connection and the system has had to rely on this error correction to ensure that the data being read/written is correct. (For casual readers, please don't assume this is correct - it's just my impression from reading.) Generally these error counts don't change unless I've opened up the case and replaced a hard disk or perhaps during a parity check. So I assume vibration or movement has loosened something. But after a few warnings the error count doesn't increase. I've taken a look at the various SATA connectors and can't see anything obviously wrong. So now I just ignore the errors unless the error count increases consistently. (My server runs 24x7.) Think this might have happened once or twice, but not recently. And I think I "fixed" it by jiggling SATA connectors or just swapping them.

I'd be interested to know if anyone else has Kohya running successfully in the docker. I ended up going back to running everything natively in my Win11 VM, largely because I don't know enough to go into the docker and do all the fiddling around that the various packages always seem to require. But while I can natively run Automatic1111 and ComfyUI, native (Windows) Kohya constantly gives me problems. I've only been able to run it successfully under Pinokio.

Diagnostics attached. tower-diagnostics-20240205-1109.zip

I have an 8TB disk which has reported 93 read errors during a (now paused) parity check. I have a 16TB pre-cleared disk on hand to replace it. Can I just cancel the parity check and replace the 8TB disk with the larger one? (My dual parity drives are 16TB). I think the answer is "yes", but just want to check in case I mess up something.

So I did a little test: stopped docker containers and VM. Set the graphics card in the VM to the Nvidia card (no VFIO binding) and then booted up the VM. It seemed to be running but I couldn't connect via RDP. So I thought about stopping the Nvidia plug-in. At that point I thought it might be a good idea to see what the the Nvidia plug-in is meant to do, and went to the support thread. I didn't understand everything, but in the first post there was this warning: "Please be sure to never use one card for a VM and also in docker containers (your server will hard lock if it's used in a VM and then something wants to use it in a Container)." Looks like my idea is horribly risky, so I've abandoned it.

Have you tried Guacamole? Don't know if it's fast enough for your purposes and whether it will do everything you want, but I found it OK and very easy to set up. (But in the end I went with Tailscale because I had other devices that I wanted to connect to in a consistent way.) Have never tried Parsec, so can't compare.

In my server I have an RTX 4070 graphics card, which is used by some of my docker containers. I also have a Win11 VM which just uses the default Virtual graphics card. The graphics card is not bound to VFIO, and I have the nVidia plug-in installed (which I understand is needed for the docker containers to use the GPU). I understand that I can't have the VM and the containers using the GPU at the same time (that would of course be ideal). As the next best alternative, I'd like to be able to switch the usage between VM and containers as needed. What I envisage is, say, starting the server with the docker containers active and able to access the GPU. Then, if I want to use the VM with GPU access, I would shut down the containers and then start up the VM. If I then want to go back to using the containers, I shut down the VM and start the containers. My thinking is that in order to be able to do this, I have to assign the GPU to the VM, but not bind the GPU to VFIO. Is there anything else I need to do? For example, would I also have to enable/disable the nVidia plug-in each time? (I know this sort of question has been asked quite a few times. I've read various threads about GPU passthrough, but most seemed to assume this also meant VFIO binding - maybe "passthrough" does, but I wasn't sure. Also, any mention of the nVidia plug-in seemed to refer to older versions.) On the risk side, assuming I can do this switching, what happens if I forget to shutdown the VM/containers and run them simultaneously? Will there just be no response or will something crash?

I tried a few more things: I installed FaceFusion in a separate container, but using the same /config folder. It worked fine, with ComfyUI running as well. I tried reinstalling Kohya again, separate container, same /config, but got the same endless loop. I installed Kohya again, but this time in a different /config folder (/mnt/user/appdata/kohya instead of /mnt/user/appdata/stable-diffusion for the others). Installation went smoothly, I saw the Torch bit looking OK in the log after the nVidia toolkit was detected: 23:32:46-955597 INFO Version: v22.5.0 23:32:46-959892 INFO nVidia toolkit detected 23:32:50-012332 INFO Torch 2.1.2+cu121 23:32:50-016632 INFO Torch backend: nVidia CUDA 12.1 cuDNN 8902 23:32:50-026714 INFO Torch detected GPU: NVIDIA GeForce RTX 4070 VRAM 12009 Arch (8, 9) Cores 46 23:32:50-027610 INFO Verifying modules installation status from /config/70-kohya/kohya_ss/requirements_linux.txt... 23:32:50-028561 INFO Installing package: xformers==0.0.21 bitsandbytes==0.41.1 The GUI came up fine (I didn't run anything). Then I shut down the container and restarted. This time I was back in the endless loop, with the same "Could not load torch" error as before, and no GUI. A partial log is attached. Kohya_log_2.txt

So now I've installed Kohya in a separate container, but using the same folder /mnt/user/appdata/stable-diffusion for /config. As expected, I ended up with a 70-kohya sub-folder there, along with my existing 05-comfy-ui folder. On initial launch it seemed OK; the interface (assigned to port 9070; ComfyUI is on 9000) came up fine, but I didn't run anything. But when I restarted after adding some paths to the template for the outputs and model folders, it went into a loop in the log file and the web UI never came up. I thought it might be a problem with ComyUI running at the same time, so I stopped CUI and then restarted Kohya, but there was again an endless loop of messages in the log file. I couldn't find it saved anywhere, but I grabbed what I could from the log after I stopped the container. That's in the attachment. I saw the log (but keep in mind I'm not a Python person) that "library" is installing and uninstalling, and I don't know whether the hyphen makes a difference. The log also says there's no venv folder, but I do have a venv folder under 70-kohya. However, ComfyUI still runs OK. I'll try installing maybe FaceFusion or A1111 in another container and see how that goes. Kohya_log.txt

Let's say in the GUI Docker tab I click on the stable-diffusion docker and click Edit to change the existing template. If I just change the WEBUI_VERSION and install Kohya in addition to my existing ComfyUI installation (done via the container), I assume this means I have both ComfyUI and Kohya in the same container. When the container starts, which UI starts running? And if I don't change the WebUI port, which UI will I see when I open the default GUI address in the browser? I might have misunderstood what you said. It seems like it's better to run multiple containers, with each one controlling the startup of a particular GUI. (I assume they can share the same /config folder.) Otherwise I don't know how I can choose whether to start Kohya or ComfyUI when they're both in the same container. I would have tried just editing the template, but I have spent some time getting ComfyUI running the same way I had it before in my VM, so I'm a little scared of borking the whole thing.

Thanks, I'm trying this out now. I have (05) ComfyUI installed. If I want to try something else like (70) Kohya, do I just edit the template and select a different WebUI port and WEBUI_VERSION 70 and apply? Or would this require a separate container? For ComfyUI I see there's a parameters.txt file that allows you to set the output directory, which defaults to a path in /config, and models are also stored in /config. In my case /config is actually a folder in the appdata share on an NVME cache drive. This drive is likely going to run out of space quickly if I store models and output there. I know that ComfyUI has a yaml file that lets you point to another folder for models. So I can put models somewhere else and point ComfyUI at that location. How does the docker container use the models folder that it sets up? Will all installed UIs automatically look there in addition to anything set up by each UI? And if so, can I move that models folder somewhere else? I would prefer not to move my appdata share. I suppose I could set up the stable diffusion config folder outside appdata, but that doesn't seem right, although I can't think of a good reason why it would be wrong. (Maybe appdata backup would be a bit of a problem?) I've tried pinokio (browser based front-end for various stable diffusion UIs) in Windows and it seems to be a little like this container - lets you download and install various UIs, but documentation is a bit sparse.

In the Docker template for Immich for Repository I used ghcr.io/imagegenius/immich:version-v1.90.1. I think this general approach of entering a specific version other than "latest" works for all docker containers if you need to roll back. You do need to make sure you find the exact numbering of the version; usually that can be found in the repository for that container. Unfortunately I now can't recreate where I found that particular version number. (Currently I'm not using Immich because it seems to be under heavy development and too many things were breaking for me when I updated it. Not a criticism - development is great; but it's not reliable enough to roll out to the family.)

I currently run Comfyui and A1111 on a Win11 VM (with GPU passthrough) on an unRAID server. All output and models etc. are kept in the VM (nothing on the array). I think using this docker might be a cleaner solution. My main concern is whether there would be a performance hit. I have a 12GB VRAM Nvidia 4070. Theoretically (or practically) is there a significant downside/upside in moving to the docker? If I do the migration, I'm thinking of following these steps: 1. Remove GPU binding 2. Reboot 3. Start Win11 VM and make sure it's running OK without GPU 3. Install this Stable Diffusion docker 4. Install SD GUIs and organise folder mappings 5. Go into Win11 VM and copy the necessary files to the corresponding docker mapped folders (I'm guessing mainly json workflows, models, custom nodes, extensions, embedding, LoRAs). 6. Might have to fix some model locations in the workflows, but that shouldn't be a big problem. Any suggestions or advice on this?