stottle

The changes to add the add to Community Applications went live this morning. I just pushed a bunch of changes to the image. It now starts with two example notebooks * chat_example.ipynb provides a running chat example * diffusion_example.ipynb provides (wait for it....) a running stable-diffusion example It currently requires an NVIDIA GPU

Overview: Support for the JupyterLabNN (Neural Network) docker image from the bstottle repo Application: JupyterLab - https://jupyter.org DockerHub: https://hub.docker.com/r/bstottle42/python_base/ GitHub: https://github.com/bstottle/python_base There are two goals for this container 1) Provide an interactive (via Jupyter notebooks) python environment set up for evaluating AI/ML applications including Large Language Models (LLMs). 2) Provide a common _base_ image for LLM user interfaces like InvokeAI or text-generation-webui that leverage python for the actual processing.

I didn't see an answer posted, but per a post in another netdata support thread, the nvidia-smi support was moved from python.d.conf to go.d.conf (and set to "no" instead of "yes"). I was able to get Nvidia visualizations by changing (in go.conf.d) # nvidia_smi: no to nvidia_smi: yes I needed to add the parameter/value NVIDIA_VISIBLE_DEVICES / 'all' to the docker template and append `--runtime=nvidia` to Extra Parameters as well. I used the following to create/edit the config file (from the post linked above):

Ok. That does make the folders accessible. But seems only slightly better than mounting from /mnt/cache/<share>.

Sorry if my original post wasn't clear. I have 7 shares, and with exclusive shares setting enabled, 3 of them became exclusive shares. Exclusive: appdata, media, pictures Non-exclusive: domains, isos, system, zfs-backup. In the top image (binhex-krusader), the left side of the window shows a folder view of /mnt/user exposed to the docker (all of the shares). The three exclusive shares are shown with a different icon and are not accessible as folders in the docker container. The non-exclusive folders are accessible and behave as expected. For the Ubuntu VM, the situation is similar. I manually expose /mnt/user to the VM, but only the non-exclusive folders show as folders, the exclusive ones have an icon with an "X" and don't work (in the bottom image). I've tried both 9p and virtiofs and had the same results. Note - I don't expect any files to move automatically, I was expecting to be able to use the docker/VM to test if the write speed was improved (via drag-n-drop in the systems) skipping FUSE. Hopefully this is clearer? Are you successfully using exclusive shares? If so, can you describe what you are doing that works?

I didn't have an issue getting shares to be recognized as "Exclusive Access". I had trouble with both dockers and the VM being able to use the shares once they were set as exclusive access. Was that not clear from the screenshots?

Alrighty. `Permit exclusive shares: No`

Is no on else having similar issues? If there is a fix posted somewhere else, can someone share a link?

Hi, I've enabled exclusive shares and I have several shares on a ZFS cache pool that show as having exclusive access. Looking at /mnt/user/ in terminal, I do see three folders that are symlinks. From this, I believe my configuration is correct and working, with `appdata` being one of the exclusive access shares. The limited dockers I've tested all seem to start and run. But results are mixed: * `code-server` is able to open exclusive and non-exclusive share folders * `binhex-krusader` set per spaceinvader1's video - only non-exclusive shares are accessible (note the last 3 folders on the left are grayed out) * Ubuntu VM - only non-exclusive shares are accessible via VM mount Here is the section from the VM config: Here is the fstab config and folder view of the mount showing the 3 exclusive folders as unaccessible: It seems like different ways to access the shares are hit or miss. Are these bugs? Something I need to change in the config? Thanks!

What operations are safe/non-destructive? Are all operations from this docker safe, as in no damage to existing content? The first topic says "Disk Drives with platters are benchmarked by reading the drive at certain percentages", but "Solid State drives are benchmarked by writing large files to the device and then reading them back". I have two NVMe drives pooled as my cache, which makes me think testing them individually could corrupt the data on the pool. Is it correct to assume benchmarking spinner disks is safe in a running server, but testing _pooled_ solid state drives is not? If only some tests are safe, what about testing controllers? Apologies if this has been answered before - I read the first 4 pages of topics and searched for "safe" and didn't find anything that answered my question. Thanks!

Looks like it was a temporary issue that resolved itself. Tried a new USB with clean 6.12.6 and that worked fine. Was installing 6.12.5 (just to check), and tried the original USB drive while that was installing and networking worked with it as well. I had tried resetting various parts of the network yesterday before posting this topic - it hadn't helped. Today without actively changing anything, everything was working. While I'd love to say I knew what root cause was, I'm good with a working system. Thanks for the help!

Here are the diagnostics. supertower-diagnostics-20231201-0829.zip

I have two Unraid servers (which have been working for a while, until today) - they have different IP addresses and different names. There is another topic with the same title and a solution related to NPM - which I have been messing with, but I don't have the same message in the logs and I'm having issues even if the array is stopped. With one array powered off and the other booted into the gui, but with the array stopped, I see the powered on server on Unraid connect (2nd on shows off). With a display/keyboard/mouse directly connected to the server, I see the GUI. But the device isn't showing on the network and Unraid connect can't connect to it. the ifconfig command shows no init address for bond0, eth0, eth1, and x.x.x.142 for br0. I don't know what ifconfig showed before I was having issues, though. It shows as x.x.x.142 on the network from Google Home (I have Nest mesh wifi) which lists wired and wifi devices connected, but pinging x.x.x.142 returns timeout and I can't connect to the GUI from local web address either. I didn't change any hardware configs, but I was messing with NPM and CloudFlare tunnels while also updating both machines from 6.12.4 to 6.12.5. I was trying to change the NPM config and restart when it failed because the IP was taken, and it looks like for a while the Unraid server changed its IP address from x.x.x.142 to x.x.x.2 (the static address I assigned to NPM) (maybe?). The other server, which isn't running NPM, seemed to have the same connection issues with 6.12.5. Please help!

Hi, I've got AdGuard running in a docker container which lets me add a filter to an IP address and give it my domain name (DNS rewrite is another term for this I believe). The intent is to be able to use the same urls from my lan as externally, where the ip address I use would be the IP for the reverse router. The issue is that NginxProxyManager seems to require 8080 for http and 4443 for https, not the default 80/443. This means even if I set a static IP for the docker, urls won't work. If I try to port forward 80/443, that conflicts with the webUI. The solution I've been pondering is creating an additional host IP address, and then port forwarding to that. Pipework seems to be an option, or manually creating an interface alias with ifconfig. Seems like Settings->Network Settings->Routing Table could help as well. Right now I'm using a cloud flare tunnel (wildcard) and not port forwarding, but ideally the ip would be something I could port forward to if desired (which I think means needing a Mac address). Is there a recommended approach? Other alternatives to consider? TIA

I've got Seagate SAS drives and an LSI HBA. Sounds like there are variations of Seagate + controllers that don't work, but the UI/log seems to show my drives spinning down. But they almost immediately spin back up. Before I installed this plugin, requesting a spin down of one of the SAS drives appeared to be a no-op (the green circle would not change). After installation (actually setting the spin down to never, rebooting, setting to 1 hr, rebooting, and then installing the plugin) now shows a spinner, then gray for the a drive, but quickly spins back up. The log simply says Aug 13 12:17:40 SuperTower emhttpd: spinning down /dev/sdd Aug 13 12:17:40 SuperTower SAS Assist v2022.08.02: Spinning down device /dev/sdd Aug 13 12:17:57 SuperTower emhttpd: read SMART /dev/sdd I have turned off docker and VMs, and have basically no data on these drives yet so nothing should be trying to access the drive other than Unraid itself. I'm on 6.12.3 if that matters. Any suggestions for fixing this or troubleshooting?

1) USB port - it is in the same 2.0 port it has been in for a very long time, with no issues until now. 2) Re-format - what is on the flash drive vs. elsewhere on the system? Are the flash contents backed up somewhere so I can replace them if I re-format the flash drive? FWIW - The UI came up (although in what looked like a daytime theme instead of the usual nighttime theme), and I could start the array so most functionality seems to be working. Root of this question is if/how I can keep my system intact if I re-format the flash drive. 3) Replace - same as #2, but with `Root of this question is if/how I can keep my system intact if I re-formatreplace the flash drive.` I appreciate the help

I upgraded from 6.9.3 to 6.11.5 yesterday. Everything I tested (mostly plex and a windows VM) looked to be working fine. Then I started getting the subject message on the UI. The requested diagnostics are attached. If it matters, I was messing with a nextcloud docker container that I haven't touched in a while, and had run Settings->Management Access->Provision to try to get rid of ssh warnings in accessing the UI when I started getting this error. What are the recommended steps here? TIA tower2-diagnostics-20221225-1146.zip

The error turned out to be a mismatch in ports between the two routers (mixing which was internal vs. external). Also, to the earlier person who mentioned still getting "insecure" messages due to having staging set to `true` - thanks, I hit that as well.

I've followed Spaceinvaderone's video for setting up SWAG, but the docker container is giving an error: Requesting a certificate for <mySubDomain>.duckdns.org Certbot failed to authenticate some domains (authenticator: standalone). The Certificate Authority reported these problems: Domain: <mySubDomain>.duckdns.org Type: unauthorized Detail: Invalid response from http://<mySubDomain>.duckdns.org/.well-known/acme-challenge/U9o-N70woR3z5jnFl0cEVPWd711PJT8SAqRPiZLYAXc [<My IP>]: "<html>\r\n<head><title>404 Not Found</title></head>\r\n<body>\r\n<center><h1>404 Not Found</h1></center>\r\n<hr><center>nginx</center>\r\n" Hint: The Certificate Authority failed to download the challenge files from the temporary standalone webserver started by Certbot on port 80. Ensure that the listed domains point to this machine and that it can accept inbound connections from the internet. Some challenges have failed. I have two gateways, AT&T for ISP and a Google WiFi mesh, but I believe I have the port forwarding correct. Two reasons for this. 1) I can see my Plex server, so the two hop forwarding to that container is working 2) I was getting timeout errors in the log, but those have now changed to this unauthorized/404 error. For SWAG, I am have AT&T forward 80 and 443 directly (the only option I saw), and Google changing the ports to 180 and 1443. SWAG is set up for 180 and 1443. I'm trying to get http auth working as that seemed like the best place to start. I need to understand the other options better, too. Any tips for debugging?

Thanks for all of the work here. I've got nextcloud/letsencrypt working with duckdns, which I wouldn't have tried without the support here and tutorials. One annoyance - is there an easy way to get unset urls (https://mydomain.duckdns.org/random_garbage) to map to 404 instead of the default "Welcome to our server?" Google searches for 404 and "welcome to our server" don't help...

arch - thanks for the great write-up. I hit a snag where the controller I wanted to pass was in an IOMMU group that had other devices I didn't want to pass to the VM, though. Since this is a PCIe card, I didn't expect this. I found this blog post about IOMMU groups really helpful, specifically After a few tries, I found a slot that put the usb controller and the GPU I want to pass in the same group. So I don't need to try the PCIe ACS override. Maybe it is worth addressing what to do if you don't have nice groups in your first post? Also, your group shell script would look pretty good in that first post! And again, thanks!

Is anybody using docker compose? Are there any plans to integrate it with unRAID?

Maybe this should be stickied?

Nice! I will have to give that a shot.