spali

That was the missing click I didn't found. Thank you.

Hi all I may have killed my supermicro BMC (ipmi) and potentially 2 nvme's of a btrfs mirror pool. I slipped installing a GPU on the board and I think I killed BMC (ipmi) on my board and also potentially my 2 nvme's of my mirrored btrfs pool. The problem with the BMC seems to be known with that board... as some resistors seem to be more than usual fragile near the pci slots. My nvme's were near too.. I can boot (even no screen due missing integrated GPU and doesn't pick up external gpu) and no remote ipmi ... So no access to the bios to check the settings.... Unraid doesn't find the disks, even for one I get the dmesg output: [ 62.784935] nvme nvme0: I/O tag 17 (0011) QID 0 timeout, disable controller [ 62.787070] nvme nvme0: Device not ready; aborting shutdown, CSTS=0x1 [ 62.792699] nvme 0000:41:00.0: probe with driver nvme failed with error -4 I already tried the disks with NVME to PCI adapter.. but same result, so I don't think it's the nvme connectors on the board. And also I can't see the disks if I put it in a windows PC. They show no obvious damage... but I start to believe I got the jackpot by killing both at the same time somehow... but one seems to be at least somehow not completely dead. So have two problems now: I hoped to resurrect the data after last backup from at least one. Maybe someone has an idea... regarding the dmesg errors, I found some kernel parameters to try... but not sure if that helps... as they seem most used in case of disconnects not half dead disks... nvme_core.default_ps_max_latency_us=0 pcie_aspm=off pcie_port_pm=off Also I can't start the array without the pool. If I'm not wrong, I should be able to check the "Yes, I want to do this" option and start the array. If one disk would be somehow resurrected, I assume I can access the data to copy... so wouldn't care to rebuild the pool with new disks anyway. Or reimport the pool if possible from one of the resurrected disks. Even with "Yes, I want to do this" option, I get "Wrong pool state" - "too many wrong or missing devices". But I can't remove the pool nor unassign the missing disks. update: as expected... booting with the found kernel arguments didn't change anything: # dmesg | grep nvme [ 0.000000] Command line: BOOT_IMAGE=/bzimage isolcpus=18-23,42-47 initrd=/bzroot nvme_core.default_ps_max_latency_us=0 pcie_aspm=off pcie_port_pm=off [ 0.163217] Kernel command line: BOOT_IMAGE=/bzimage isolcpus=18-23,42-47 initrd=/bzroot nvme_core.default_ps_max_latency_us=0 pcie_aspm=off pcie_port_pm=off [ 2.291209] nvme nvme0: pci function 0000:41:00.0 [ 62.784860] nvme nvme0: I/O tag 9 (0009) QID 0 timeout, disable controller [ 62.787002] nvme nvme0: Device not ready; aborting shutdown, CSTS=0x1 [ 62.792852] nvme 0000:41:00.0: probe with driver nvme failed with error -4 update2: also tried both of iommu=soft amd_iommu=off also no change unfortunatly. host01-diagnostics-20260321-2051.zip

Have same errors during rclone backup over samba to a second unraid server (using unassigned drive mount). Accessing the mount in any way often hangs during seeing these "general protection fault". I specifically notice this during backup of container data files of "loki" and "tdarr". For "loki" its especially the files in "/loki/chunks". For "tdarr" the files in /app/server/Tdarr/DB2/ResJSONDB. Don't know much about the specifics of those files, but both directories contains tons of highly volatile files that changes all the time. But I have it even if I stop the containers. So it doesn't seem to be the problem of only accessing "in use" files (rclone usually throws just an error and continues, this worked in the past without problems.). Maybe a combination of the amount of files and maybe (guess) shfs which screws it up so it's also a problem when containers are stopped. Sometimes find some ".fuse_hidden..." files. Usually the system just reacts slowly during accessing these files and samba mount related access is slow or with some freezes. But as soon as I force kill all backup processes the system seems to recover. Can't say for sure when this started, currently on 6.11.5. When I exclude these directories I get a backup run without these "general protection fault" entries in the log. Preparing for upgrade to 6.12 ... and will try the "Exclusive Shares" for container files... this may help and at least eliminates the factor of the fuse layer on that files.

Not till now as I know. At least not in docker settings page on version: 6.11.5.

I think the main problem is that we have at least two use-cases for people using docker compose. 1. the people which just want no special features, but reuse compose configs. 2. the people which want most, if not all compose features, which usually also create their own compose files. I count myself to the second group. So that's why this kind of "compose - dev - container" works perfectly for me. What I don't get is what's the plan with the step with the 6.10 feature of the container labels for just icons and webgui url. I don't know their thoughts and plans, but I would be consequent and just replace the templates with compose-files. They would only need to support the features they want in the GUI and just but unraid specific stuff in labels (as with the comming changes for icon and webgui url). Even if they would decide to support only one container in the GUI per compose file. It would be a lot more compatible with custom compose files, especially that (at least my impression) a lot of users here use them. It would also allow for people in the second group to to at least basic stuff in the GUI with their custom compose files. I see the point with the storage location for volumes... but that's something a "power-user" anyway has to deal with. A so called compose based template would just predefine the unraid paths, as the current templates do. But it would be an easy adaption from a compose file that I copy&paste from the internet. Maybe the labels they implement in 6.10 is only a small step in their plan towards this direction, who knows....

In the compose you can change the variables in the compose file for the versions. Just make sure to use the right notation, because they are github tags/releases some of them with "v" prefixed. And the DOCKER_VERSION argument is also referenced for the docker image used to pick the docker binary. But I'm sure if primeval_god decides to integrate it in his plugin, he will make it more polished and configurable.

@primeval_god suggestion for "GUI". I just prepped a dockerdev container for my unraid compose management as a supplement for your plugin. Why not include just a vscode server container for editing the files. I have a working compose file that setups a vscode server container based on linuxservers/code-server with docker, docker compose with same version as you provide in the plugin and some suggar. Features: vscode extension: docker support for managing/debug docker via vscode docker explorer. support for compose up down commands. synatx highlighting and help for compose and docker files vscode extension: dotenv for all the people using .env files bash completion in the vscode terminal for docker and docker-compose I'm happy if you would integrate that with your plugin, I assume you would have less effort to bring this up and running and support with more features than self dev a gui. Maybe it needs a bit of cleanup und convert to a template but should be almost 1:1 mapping. Explanation: The docker file builds the sugar into the code-server image, they have a "mod" for integrating docker into their images, but hard to maintain with custom version etc. The settings.json of the workspace is currently manually created after container is running... but could be easy integrated into the docker image as base. And maybe I missed some docker command settings for all possible cases. and currently also hardcoded for using .env file in these commands (couldn't get it to pick it up automatically). Also the image needs some cleanups like removing apt cache etc. to make it smaller. dockerfile for codeserver with sugar: docker-compose: vscode workspace settings file:

Just for information if someone stumble over this post. I struggled with it and a lot of posts found by google actual gave the me the impression of there is a deeper technical problem or not supported well (especially with the Unifi switches as in my case). Also the "Warning: No 802.3ad response from the link partner for any adapters in the bond" message lead to the wrong direction. I started to believe it's just not working. But the truth here is... it actually works and that without a problem. But took me a while to find this post about it's just a GUI bug You can prove it for yourself with the following commands. Executing in the shell (change the bond number to your bond) to see if the bond is active: # show current active mode cat /sys/class/net/bond0/bonding/mode # show more details abount the bond cat /proc/net/bonding/bond0 In my case also the switch reports the LACP bond as working and up. @Moderators Maybe this should be pinned till it is fixed (I assume in 6.10 , I found tons of posts without a solution (except suggestions to use multi channel or another bond mode), but none except this after hours of reading such posts and try and error.

Long time ago (years), played with tuners... but also new at unraid. My plan was to use my existing HDHomerun with TVHeadend and Emby later in the mix. Emby is new for me... just read in the past about it but never got time to check it out. I can only say, that TVHeadend works perfectly fine with HDHomerun. Also tons of reading about it by googling. I used it with Kodi with the TVHeadend plugin. But as soon I'm happy with my unraid setup and moved everything to it, I will start from scratch with that topic.

Hi I know, first post but would love to learve my two cents here for this topic. To be most flexible, especially with docker-compose in mind and the coming label support in 6.10. I also already using labels for my CI/CD system and traefik as proxy. I would suggest to use a label like dockerFolder.folderId as default. i.e. when it exist's automatically create a folder based on it. But also provider a functionality to define custom labels... i.e. allow to make a folder with all containers that contains a label (optional with a specific value). Some use cases I have: create a folder which contains every container that has traefik.enable=true and also one folder with traefik.enable=false. This would allow me to have a nice overview of all exposed and non exposed containers. I have my local CI/CD system with jenkins that automatically provisions it's agents (docker containers in unraid). They all have specific labels for their purpose of building. Or could also be just a container where a unraid backup script is running. Here I also use specific labels. Consequentially this would allow containers to be in multiple folders. But also adds a huge flexibility if you want to reuse existing labels without to support them specifically in the plugin. As a simple implementation proposal, I would suggest to let the user configure multiple label=regexp configurations on the container. Similar as now it's possible to have one regexp on names, but just with a predefined label as key.