JesterEE

Recently after installing Unraid 6.11.0 (and subsequently 6.11.1 in hopes of fixing this issue) my server started reporting kernel NULL pointer dereferencing after about 1-2 days of uptime. This causes: The WebUI to become unresponsive (loading indefinitely after "successful login") Some (but not all) on my dockers to become unresponsive Dockers like plex and the *arrs seem to stay alive Others become unresponsive via their web interfaces but are reported as up via a 'docker container ls' check. Non-functional powerdown and poweroff commands (in attempt to reboot without parity check) It does not seem to impact: Operation of a VM I have running before, during, and after the error is reported (I'm typing on that VM right now and I had this issue pop up earlier this AM). Connection to the server via SSH For full transparency, I did upgrade my server's hardware (CPU and added RAM) 2 weeks before the 6.11.0 upgrade when operating on 6.10.3. I ran memtest86 on all the RAM for 3 passes after installation so I'm pretty sure that's OK. And 2 weeks of stable Unraid 6.10.3 operation leads me to believe this is unrelated to my hardware upgrade, but I cannot be 100% sure. Here is the relevant section of the syslog. Diagnostics attached. I have other diagnostics showing the same issue on both 6.11.0 and 6.11.1. Any help from the gurus appreciated! 🙏

I have the same issue on my reverse proxied Unraid dashboard, but only on a certain Windows 10 device. The browser loading the content doesn't seem to matter (tried Google Chrome and Microsoft Edge) so I'm thinking its some group policy or something; maybe even an IT security safeguard (i.e. blocker). I looked at the network traffic in the Chrome Debugger and see all the WebSocket traffic coming in; none of which is blocked outright by the browser. But, I am getting errors in the console that look like this: So, something is not letting the connections through. I'd like to know how to fix this too, but I have no idea what would be causing this issue outside the browser. My best guess ... an IT deployed "safeguard". @FlyingTexan, have you tried on a different device?

For anyone looking for a currently available FL1100 based USB Controller that works with Unraid, Linux and Windows 10 VMs (haven't tested a macOS VM) check out this one from YEELIYA. I haven't tested throughput to see what I can get out of it in a real-world test to really push the PCI-E 1x bus, but I have no issues with a keyboard, mouse, webcam and 2 USB sticks working simultaneously.

@JorgeB certainly did a lot of the heavy lifting on this one. I have referenced that post multiple times when making my own HBA considerations a few years ago, and it's probably one of the best data driven studies on the subject. While making your decision, just remember what you're controlling; likely spinning rust (HDDs). Most of these drives you're likely going to use for your data are moderate speeds (5400 RPM) and have an I/O limit of ~100 MB/s in real world application. So, while it may be great that your fancy PCIe 3.0 card doubles performance against the PCIe 2.0 counterpart ... do you really need it given the significantly higher price tag? SDD array, likely yes ... HDD array, probably not. Also be cognizant of the PCI Express lanes your CPU and motherboard combination offer for the application. Read the fine print on the motherboard specification to make sure you'll be able to use all the lanes that your card offers. It's not uncommon for the lower priority 16x slots to only be able to source 8x/4x speeds given a consumer grade configuration.

This USB pass-through write-up is still relevant and I performed it today on my ASUS ROG Strix X570-E on the latest BIOS build (Version 4021) which appears to have an identical architecture as the GIGABYTE X570 AORUS. One thing to note ... the internal USB headers are also attached to the same controller we are able to pass through. So, if you need to run an internal peripheral and want to keep it attached to the Unraid host (e.g. a fan controller, LED controller, etc.) you're SOL! 😡 There are 2 internal USB2 headers on the board(s), but they are both attached to the same controller. I wish they split them between the controllers, but from a board design and layout standpoint, I understand why they didn't. -JesterEE

Something changed in the last update. Not sure what, but this wasn't happening before. I just checked my docker and after a few days or running, and streetmerchant was eating all my RAM. I started limiting the RAM consumption of the docker now. We'll see how this works out.

@kiowa2005 This project is updated very regularly. Updating links, adding features, fixing bugs, etc. Is there a way to structure the dockerfile such that users can force an update so the image can pull from the github repo and run the latest code?

@zepheris @kiowa2005 Using the built in proxy support also works: https://jef.codes/streetmerchant/reference/proxy/ You just have to map the files corectly to the docker: For example: -v '/mnt/user/appdata/streetmerchant/global.proxies':'/app/global.proxies':'ro' I use this for amazon so I don't hit the servers too hard and start forcing CAPTCHA requests. -JesterEE

Oh there are definitely other cards...I'll get back to you in a while with a list of links. Are you in the dev team? Do you want me to track this elsewhere (github)?

@kiowa2005 Is this container pulling from master or the latest stable release?

I haven't tried it, but the standard way to pass username and password fields to a proxy is: protocol://user:[email protected]:port so for example https://username:[email protected] (you probably don't need the port if it's the standard port for the protocol)

@kiowa2005 Thanks for this! Got this working today with a Telegram channel and bot I set up. Took about an hour and now I'm not getting stupid PS5 notifications from those "other" stock notifiers pushing to Discord/Twitter/Telegram 🤩. I was monitoring my log file and even though I specified asus as a store endpoint it doesn't seem to be being checked for 30 series cards. I know this has nothing to do with the container, but has anyone else seen that on their installs? I'm also getting the CAPTCHA for amazon. The store as a whole seems to 'greylist' you ... so PAGE_BACKOFF_MIN doesn't seem very helpful as you just nail it with requests for another product. For now, I just turned off amazon ... I'll play with it when I'm not on there 💩 list. It would be nice if there were different back off periods for different stores and store product pages ... but again, this is not the containers fault.

@bonienl I started using this plugin on a different array the other day and I think I found a bug with the reporting mechanism for the Export being up-to-date with the Build. When mid-Build, I accidently hit the Export button the the webUI with all the disks selected. The build continued, but an export was also done on all the devices in my array with the current files where hashes were computed. I figured, these would almost certainly contain incomplete data, but I'd make a new export when the build completed the next day. Before I did however, I wanted to see if the Export was incomplete AND the webUI reported that the state was stale. When checking the export files, I was correct that the data was incomplete and not all the hashes were in the export files. BUT, the webUI greeted me with all green check marks on the export status for each device. This left me with a bigger question. How does the plugin determine if the build and export are up-to-date?

That's an interesting find @keefd! Probably one of the best solutions for a modern server build where you want to incorporate some "average sized" PCI-E cards without losing space for HDD cage mounting. I do agree though, more than I'd want to spend (and that doesn't even cover the HDD cages). I've never heard of the manufacturer before ... I'd be interested in a build quality review ... So it looks like a pretty impressive case, with average or maybe a little better than average build quality. I think they priced themselves out of the market though. Fractal Design just released the Meshify 2 XL at $180 only a few months after this chassis' release. It's basically the same size and construction as the Define 7XL but it can breath instead of roast and is $20 cheaper. Even when you buy the additional trays to complete the build, you'll still be under $300 for a loaded chassis (comes with 6 trays, need 10 more for the stack [10 trays / 2 trays per package = 5 packs * $14 = $70 + $180 = $250 for 16x 3.5" drive locations and 4x 2.5" locations]). If you were set on hot swappable drives though, the Anidees AI Crystal XL PRO may be the one to beat! I personally just bought a Meshify 2 XL ... and as I build in it, I am very impressed! Still short of the 20x 3.5" drives possible with 12x 5.25" slots, but I'm happy with this case for my needs. Really still want that backplane though 😰. -JesterEE

Follow-up from my last post. TL;DR; APC Sucks So I got an APC RJ45 to RS232 (Serial) cable and a RS232 to USB adapter to do the firmware upgrade needed to hopefully get MODBUS over USB enabled on my UPS. Well, in APC's infinite wisdom, they apparently have different firmware ceilings for different units with the SAME SKU. Looking at the APC FAQs I found this entry on updating the firmware. So, I have a SMT1500US toaster, so I used Firmware Update Wizard 4.1.1. From the listing of firmware updates, my unit could be an SMT 4.5G with Micro-Link 17 or 18. My unit was at firmware version 5 and didn't report a Micro-Link version so I had no idea which to expect. Plugged in the cabling, launched the Wizard, and it proceeds recognize my unit and suggest an update to firmware version 7.1. HOSED! 😡 Since the thread I linked in the last post suggests that I need firmware version 9.1 to use MODBUS over USB, I'm now SOL. Who releases a 2 different products that look exactly the same without a way of deciphering which is which!? I even contacted APC support 2 weeks ago and they couldn't even tell me which I had based on my device serial number. I have to say, as an engineer myself, this is sloppy, sloppy work on APCs part and now I am very frustrated. -JesterEE

FINALLY! If you have a case with lots of spinners and "thermal zones" this is the product and SUPPORT we have all been waiting for! -JesterEE

Sorry to necro this post, but I was going to post a new thread about it anyway. EDIT: apcaccess result APC : 001,027,0657 DATE : 2020-07-08 12:24:13 -0400 HOSTNAME : Tower VERSION : 3.14.14 (31 May 2016) slackware UPSNAME : Tower CABLE : USB Cable DRIVER : USB UPS Driver UPSMODE : Stand Alone STARTTIME: 2020-07-08 12:23:11 -0400 MODEL : Smart-UPS 1500 STATUS : ONLINE BCHARGE : 100.0 Percent TIMELEFT : 129.0 Minutes MBATTCHG : 10 Percent MINTIMEL : 10 Minutes MAXTIME : 0 Seconds ALARMDEL : 30 Seconds BATTV : 27.0 Volts NUMXFERS : 0 TONBATT : 0 Seconds CUMONBATT: 0 Seconds XOFFBATT : N/A STATFLAG : 0x05000008 MANDATE : 2011-04-08 SERIALNO : z NOMBATTV : 24.0 Volts FIRMWARE : COM 02.1 / UPS.05.D END APC : 2020-07-08 12:24:56 -0400 Note, no LOADPCT, or NOMPOWER 😭. So, while the below is all still valid ... the issue is my UPS is not reporting what is needed for the plugin to work. Was wondering if more information was output when using MODBUS driver instead of the USB driver, but I couldn't get it to work on my unit and I see no option in the UPS physical config. Googling, I found this thread saying that I need a newer firmware to support MODBUS over USB 😭. Looks like I need to try and find a serial cable after all 😭. Sorry again for the necro when I obviously should have checked this all first. @CHBMB it looks like you were on to something. I know this was 4.5 years ago, but do you remember what was going on? Maybe it's still working for you on your current setup and you can explain what you have working. On my setup I originally plugged into a USB3 port, but after running lsusb -t, I see that even thought the port is USB3, the bus is running at USB2 speed, the device is running at USB1 speed. root@Tower:~# lsusb -t /: Bus 06.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/4p, 10000M <<<< USB3 speed /: Bus 05.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/4p, 480M <<<< USB2 speed |__ Port 1: Dev 4, If 0, Class=Human Interface Device, Driver=usbhid, 12M <<<< USB1 speed I tried another USB2 and USB3 port with no change in my load being reported or the bus/device speeds. Lastly, in my BIOS, USB XHCI Handoff was Enabled, so Unraid should be negotiating the USB speeds. For kicks, I Disabled that setting so BIOS has the obligation of doing this task. On reboot I get the same output from both the UPS and the lsusb -t command (which is interesting, but unrelated to this thread). So I don't think its a compatibility issue, but now I have no ideas. My APC Smart-UPS SMT1500US is on firmware version COM 02.1 / UPS.05.D. This is not the newest for this unit (UPS.07.1 is according to APC) but I can't upgrade my unit over USB ... only over a serial connection I don't have so ... ¯\_(ツ)_/¯.

Enhancement Request: Check for non-reduntant btrfs cache pools. I thought I was doing good making my cache redundant with a btrfs raid1 until one drive had a failure, and I couldn't mount the 2nd drive's file system. For those users that created their cache arrays in 6.7.X, this is an issue that most people are likely unaware of and would certainly cause data loss in what should be a protected event. It would be nice if this plugin would give users the heads up that something is wrong, why, and how to fix it. -JesterEE

@Frank1940 Thank you for this work. Just as relevant in 2020 with 6.8.3 as it was in 2016 with 6.2.

Maybe in the future. All the updates and support are great! Thank you! My Dashboard docker alignment still gets a little skewed using Docker Folder. I know you did some work with it, but on my dashboard, the folder contents shift to the right when the Started Only setting is selected. I think it doesn't recompute the docker folder splitting based on that option.

Quick note about this kernel build with regards to the fix for onboard audio passthrough on x570. I have been tracking the Linux kernel git source to see if this issue has been addressed on the trunk, but unfortunately, it has not. There is also the possibility that AMD would address this issue in firmware with an updated AGESA ... but the sparsely documented notes for the latest to-be-released AGESA 1.0.0.5 firmware say nothing about the on-board audio. So the if we have any hope of using the on-board audio on our boards in Unraid, a patched Unraid kernel with this patch will be required. Luckily, we have a Skitals! As a user of the NVIDIA Unraid build, I'm going to be forced to choose between my GPU in docker and my audio in my VM, which is unfortunate. It looks like I'm in the market for a USB audio DAC. 😒 EDIT: Looks like some people asked AMD to comment about the USB and audio with the AGESA 1.0.0.5, and others actually tried it ... no dice!

Enhancement Request: Can a monitor be added for the PCI Rx and Tx bus speeds reported in the nvidia-smi -q -x command? Is has been reported in the logs here as: <nvidia_smi_log> <gpu id="00000000:0A:00.0"> <pci> <tx_util>320000 KB/s</tx_util> <rx_util>3686000 KB/s</rx_util> </pci> </gpu> </nvidia_smi_log> Thanks! -JesterEE

Ya it changes the WebUI of the docker page by grouping dockers together into collapsible "folder" groups. Really, kinda handy! I'm actually hoping it gets rolled in by the Unraid devs after the bugs are ironed out 🤣. I'm not expecting a fix for ControlR as it's a plugin and not a mainstream UI feature but, just really an FYI for a known conflict.

New Unraid plugin Docker Folder breaks the parsing the ControlR app does. Dockers that are on are showing as off.