NixOS Deployments | 1240 Members | |
| NixOS Deployment tooling | 303 Servers |
| Sender | Message | Time |
|---|---|---|
| 22 Nov 2024 | ||
In reply to @khartahk:matrix.orgcan you not just boot off a USB stick or SD card | 16:11:48 | |
That's where I have the issue, the USB 3 ports on the device suddenly stopped working. The hub is visible if I do lsusb but no device is found to be attached. Then I only have 2 USB 2 ports available and the internal USB header which has the QNAP DOM module with 512MB of size. Right now I'm booting the device using a USB with Unraid on it and that work perfectly since Unraid boots to ram and does not use the drive a lot so the USB 2 speeds are not really a problem. And also the OS is quite small. I think arround 500MB or something like that. | 16:33:25 | |
| Are you sure it's not some sort of kernel bug? Have you tried booting NixOS from a USB? | 16:36:45 | |
| I tried booting the NAS using Ubuntu and there is no USB device found when plugged into the USB 3 ports, then I booted the NAS from a Ubuntu disk attached to SATA ports and connected different devices to the USB 3 port but no device was found. Also looking in the BIOS no USB drive is detected from which I could boot but if I move the drive to USB2 then it's detected and works. And I know the ports worked coz at first I was able to boot from them the same drive that is no lo longer found. | 16:41:59 | |
In reply to @khartahk:matrix.orgI'm not sure I understand the bit about issues recreating the OS. You want to be able to pull the drives and put them in a different NAS box? | 16:42:24 | |
| yes, in case I needed to migrate coz the NAS would die or something and I want data to be decoupled from the OS. | 16:44:10 | |
| Having a slice with the OS on the drives doesn't hurt - you can still put the disk into another box to recover the data. Should have backups anyway. ;) | 16:45:13 | |
| * Having a slice with the OS on the drives doesn't hurt - you can still put the disks into another box to recover the data. Should have backups anyway. ;) | 16:45:28 | |
| Since you mentioned it's a Qnap box - IIRC they're doing exactly like this. That's why they can get away with such a tiny boot device. | 16:46:27 | |
| * Since you mentioned it's a Qnap box - IIRC they're doing it exactly like this. That's why they can get away with such a tiny boot device. | 16:46:40 | |
| That is also true, I just prefer it that way. Probably coz of something I read regarding ZFS and that it need the entire drive etc ... which I know is not true but I kinda just stuck with that as it mitigates the chance of me accidentaly format the wrong drive. Especialy if the OS drive has a totaly different naming schema like nvme or something :) | 16:47:08 | |
| Fair :) | 16:47:56 | |
| At this time, it seems mostly a trade-off between requirements and time-to-working-solution. | 16:49:01 | |
| The time I have - it can wait a bit, the problem I think is I try to over-complicate things that I want to simplify. I want to move as much stuff as I can to nix because of it's declarative nature and that way I don't have to have some special backup for the config. Perhaps for this device it's easier and more cost efective to just use Unraid and have the config be backed up in another manner. Coz Unraid does everything I need: boots from usb 2 quite fast, runs in memory and does not use a lot of it so even those 2GB of ram available in here are mostly free for me to deploy some services. | 16:56:00 | |
| Side note: Since you mentioned using ZFS, IIRC that can use a lot of memory, especially when enabling deduplication. | 17:00:52 | |
| Yeah generally deduplication is not recommended unless you are in a very special circumstance | 17:02:05 | |
| Zfs without dedup eats memory because of ARC, the solution there is to tel zfs to keep minimum of X GB free, so that other software doesnt think its under constant memory pressure | 17:04:12 | |
| (ARC is the zfs equivalent of the page cache on linux) | 17:04:38 | |
| 18:10:33 | ||
| 18:47:46 | ||
| 20:14:30 | ||
| 23 Nov 2024 | ||
| 09:45:24 | ||
| Hi, I'm working on an automated update of my flake inputs and then remote switch of all my servers. I'm using a systemd service with a timer to run a script on my desktop workstation which i always leave on to update the input, test the builds using build-vm and then switching the config on the servers. I would also like to switch the config of my workstation but the switch stops the systemd service which then stops the switch, does anyone here have a similar setup and a solution to that problem? I'm thinking a cron job instead of systemd timer would maybe do it, but the wiki says it's recommended to use a systemd timer and I would prefer this for the journalctl logs. | 12:25:57 | |
| 20:48:52 | ||
| 24 Nov 2024 | ||
| 00:41:11 | ||
| 25 Nov 2024 | ||
| 07:12:52 | ||
| 07:44:04 | ||
| 08:54:54 | ||
| Hi 👋 How reliable is Nix for building release binaries targeting non-Nix systems? I try to use Nix inside my CI for building (and cross-building) release binaries. People who download such binaries are mostly non-Nix users. I'm facing issues and dilemna with both static and dynamic linking:
| 09:06:58 | |
| https://github.com/hsjobeki/nixpkgs/blob/migrate-doc-comments/pkgs/build-support/release/debian-build.nix https://noogle.dev/f/pkgs/releaseTools/debBuild https://wiki.nixos.org/wiki/Nixpkgs/Building_RPM_DEB_with_nixpkgs This one gets you pretty far in one direction, not sure if it's the direction you want to go to | 09:11:38 | |