| 12 Aug 2024 |
@brian:bmcgee.ie | In reply to @matthewcroughan:defenestrate.it BMG: if you have imported Disko into a nixosConfiguration, then a VM Test is already present for that in system.build.toplevel.installTest Good to know. makeDiskoTest is a better fit for use case in this instance https://github.com/numtide/nixos-facter/pull/73 | 16:05:12 |
@brian:bmcgee.ie | https://github.com/numtide/nixos-facter/pull/73/files#diff-35d68d8d6463519724b4aef23da56f8e50797dc42899592a9003ecb7c475cddc | 16:05:31 |
| Tanvir joined the room. | 17:10:12 |
| 13 Aug 2024 |
SigmaSquadron | Redacted or Malformed Event | 22:06:55 |
SigmaSquadron | Redacted or Malformed Event | 22:07:28 |
| 14 Aug 2024 |
| ·☽•Nameless☆•777 · ± joined the room. | 08:57:07 |
matthewcroughan | I can confirm the issue with nixos-install running out of memory still happens even with the latest disko and nix tooilng
evaluating derivation 'git+file:///home/matthew/git/nixcfg#nixosConfigurations.bpi-r4.config.syerror: build of '/nix/store/g952ci9c1fhdn61ybb71xd7yngspsbdk-homerouter-disko-images.drv' on 'ssh-ng://matthewcroughan@aarch64.nixos.community' failed: builder for '/nix/store/g952ci9c1fhdn61ybb71xd7yngspsbdk-homerouter-disko-images.drv' failed with exit code 123;
last 10 log lines:
> cp: cannot access '/nix/store/h60m1fwahjd2mv6gsg77ji3vb4gpj4dk-source/pkgs/applications/misc/tui-journal': Cannot allocate memory
> cp: cannot access '/nix/store/h60m1fwahjd2mv6gsg77ji3vb4gpj4dk-source/pkgs/applications/misc/tuir': Cannot allocate memory
> cp: cannot access '/nix/store/h60m1fwahjd2mv6gsg77ji3vb4gpj4dk-source/pkgs/applications/misc/tut': Cannot allocate memory
> cp: cannot access '/nix/store/h60m1fwahjd2mv6gsg77ji3vb4gpj4dk-source/pkgs/applications/misc/tuxclocker': Cannot allocate memory
> cp: cannot access '/nix/store/ia3ymz0pcg0w31091fb3w9jld3qa1kgp-util-linux-2.39.4-lib/share/locale/de/LC_MESSAGES': Cannot allocate memory
> cp: cannot access '/nix/store/ia3ymz0pcg0w31091fb3w9jld3qa1kgp-util-linux-2.39.4-lib/share/locale/es': Cannot allocate memory
> cp: cannot access '/nix/store/ia3ymz0pcg0w31091fb3w9jld3qa1kgp-util-linux-2.39.4-lib/share/locale/et': Cannot allocate memory
> cp: cannot access '/nix/store/kgdz4hwrdf6mir55xzmgmdnf9z75jm3l-tzdata-2024a/share/zoneinfo/right/Mexico': Cannot allocate memory
> cp: cannot access '/nix/store/l7gmwp99pwx2s20hy3hzhva0vxzpcwa9-systemd-minimal-256.2/share/dbus-1/services': Cannot allocate memory
> [ 152.515840] reboot: Power down
For full logs, run 'nix log /nix/store/g952ci9c1fhdn61ybb71xd7yngspsbdk-homerouter-disko-images.drv'.
| 17:58:10 |
matthewcroughan | All that can really be done in this scenario is trying again until it works | 17:58:30 |
matthewcroughan | It doesn't happen every time | 17:58:38 |
SigmaSquadron | Hi all. Is there a way to ensure a certain disk is processed first in a disko configuration? I have a detached-header LUKS setup that encrypts my internal disks and writes their headers to an USB device. Disko does encrypt both devices, but it processes the USB last, and it does so by deleting the headers and repartitioning. Is there a way to ensure it formats the USB before installing the headers? | 21:06:50 |
SigmaSquadron | * Hi all. Is there a way to ensure a certain disk is processed first in a disko configuration? I have a detached-header LUKS setup that encrypts my internal disks and writes their headers to an USB device. Disko does encrypt both devices, but it processes the USB last, and it does so by deleting the headers and repartitioning. Is there a way to ensure it formats the USB before installing the headers?
Sorry, I was looking for the priority option. | 21:12:26 |
SigmaSquadron | * Hi all. Is there a way to ensure a certain disk is processed first in a disko configuration? I have a detached-header LUKS setup that encrypts my internal disks and writes their headers to an USB device. Disko does encrypt both devices, but it processes the USB last, and it does so by deleting the headers and repartitioning. Is there a way to ensure it formats the USB before installing the headers?
Sorry, I was looking for the priority option.
| 21:20:39 |
SigmaSquadron | * Hi all. Is there a way to ensure a certain disk is processed first in a disko configuration? I have a detached-header LUKS setup that encrypts my internal disks and writes their headers to an USB device. Disko does encrypt both devices, but it processes the USB last, and it does so by deleting the headers and repartitioning. Is there a way to ensure it formats the USB before installing the headers?
Sorry, I was looking for the priority option.
| 21:20:49 |
SigmaSquadron | So, the priority option doesn't seem to reorder disk partitioning, only the partition creation order. | 21:22:24 |
SigmaSquadron | actually, it might not be a partitioning issue, just some disk weirdness: | 21:50:31 |
SigmaSquadron |  Download IMG_0596.png | 21:50:41 |
SigmaSquadron | i'm using this disko configuration: | 21:55:12 |
SigmaSquadron | Download partitioning.nix | 21:55:18 |
SigmaSquadron | Something is up with udev; the disk ID isn't a link to /dev/sdb. Not a disko issue though.
| 22:20:47 |
SigmaSquadron | * Something is up with udisks; the disk ID isn't a link to /dev/sdb. Not a disko issue though.
| 22:20:56 |
| 16 Aug 2024 |
| GlaringWeakness left the room. | 02:06:20 |
| 17 Aug 2024 |
| ·☽•Nameless☆•777 · ± changed their profile picture. | 08:27:42 |
pxc | oops, I'm in this channel and forgot it existed | 22:23:52 |
pxc | is anyone here having success testing disko installation on using --test-vm on macOS? | 22:25:11 |
| 18 Aug 2024 |
pxc | * is anyone here having success testing disko installation on using --vm-test on macOS? | 00:17:15 |
pxc | ok so I guess what I want to happen is for when I run nixos-anywhere ... --vim-test on a Mac, I want it to build the VM on an aarch64-linux remote builder, but then run the final test (boot the thing and whatever) by running qemu directly on the Mac.
But right now that test lives inside the nixosConfiguration, as <machine>.system.build.installTest, which means of course that it is also aarch64-linux, so that final qemu invocation is still running under my aarch64-linux builder
which sucks because that doesn't have any working qemu accelerators other than tcg
| 00:26:46 |
pxc | * ok so I guess what I want to happen is for when I run nixos-anywhere ... --vm-test on a Mac, I want it to build the VM on an aarch64-linux remote builder, but then run the final test (boot the thing and whatever) by running qemu directly on the Mac.
But right now that test lives inside the nixosConfiguration, as <machine>.system.build.installTest, which means of course that it is also aarch64-linux, so that final qemu invocation is still running under my aarch64-linux builder
which sucks because that doesn't have any working qemu accelerators other than tcg
| 00:26:53 |
pxc | for now I'm doing my local testing a little more manually, with some customized install media that mounts the repo containing my NixOS config using a VirtFS share
if I can figure it out, though, would moving disko to use pkgs.nixosTest instead of make-test-python.nix as its entrypoint for NixOS vm-tests, and then modifying nixos-anywhere.sh to support running that test locally be an agreeable change?
| 17:00:19 |
pxc | * for now I'm doing my local testing a little more manually, with some customized install media that mounts the repo containing my NixOS config using a VirtFS share and includes disko and nixos-anywhere on it
if I can figure it out, though, would moving disko to use pkgs.nixosTest instead of make-test-python.nix as its entrypoint for NixOS vm-tests, and then modifying nixos-anywhere.sh to support running that test locally be an agreeable change?
| 17:00:51 |
pxc | * for now I'm doing my local testing a little more manually, with some customized install media that mounts the repo containing my NixOS config using a VirtFS share and includes disko and nixos-anywhere on it
if I can figure it out, though, would moving disko to use pkgs.nixosTest instead of make-test-python.nix as its entrypoint for NixOS vm-tests, and then modifying nixos-anywhere.sh to support running that test locally when it is used on macOS hosts be an agreeable change?
| 17:01:40 |