| 4 Mar 2025 |
 projectinitiative | * I was also looking to do the same thing! Unfortunately in the current form, only simple single disk formatting via the mkfs bindings are supported. I created a branch to explore the more advanced use cases, and it does work (ish) for a more advanced setup.
However there are some caveats that will prevent the code from being merged into disko at this time because of the reliance on some external configurations, and a newer kernel implementation, and some testing inconsistencies. If you are interested, I can write up some basic instructions. If you are feeling like exploring yourself in the meantime, here is how I am using it to provision one of my servers:
https://github.com/ProjectInitiative/dotfiles/blob/main/systems/x86_64-linux/capstan2/default.nix
and my fork of disko:
https://github.com/ProjectInitiative/disko/blob/master/example/bcachefs-multi-disk.nix | 16:34:45 |
| projectinitiative | * I was also looking to do the same thing! Unfortunately in the current form, only simple single disk formatting via the mkfs bindings are supported. I created a branch to explore the more advanced use cases, and it does work (ish) for a more advanced setup.
However there are some caveats that will prevent the code from being merged into disko at this time because of the reliance on some external configurations, and a newer kernel implementation, and some testing inconsistencies, as well as a bug in systemd and bcachefs that doesn't properly mount UUID multi-disk filesystems on boot, so a custom systemd mount is needed (can't be used for root at the moment). If you are interested, I can write up some basic instructions. If you are feeling like exploring yourself in the meantime, here is how I am using it to provision one of my servers:
https://github.com/ProjectInitiative/dotfiles/blob/main/systems/x86_64-linux/capstan2/default.nix
and my fork of disko:
https://github.com/ProjectInitiative/disko/blob/master/example/bcachefs-multi-disk.nix | 16:36:37 |
|  lassulus changed their profile picture. | 17:48:30 |
| 9 Mar 2025 |
|  food style edible product joined the room. | 19:39:44 |
| food style edible product | I'm having a bear of a time getting KDE logging in after making a live disk with disko-install.
The issue seems to be that /home/foobar is not owned by the user, it is owned by uid 1001. The single user created is uid 1000. (Group id is also wrong, but I assume it's the same issue)
I am also using impermanence.
I have tried:
environment.persistence."/persist".directories."/home/foobar".user = "foobar"users.users.foobar.home = /home/foobarusers.users.foobar.createHome = true- home-manager `home.homeDirectory = "/home/foobar"
Any ideas? | 22:04:00 |
| 10 Mar 2025 |
|  Marie joined the room. | 15:22:37 |
| Marie | Hello, I want to add a new zfs dataset to my existing config, can I just re-rerun the disko format script, or will it destroy old stuff?
I think that should be safe, but I'm not 100% sure from looking at the generated script. | 15:24:53 |
| 11 Mar 2025 |
|  @nezia:matrix.org left the room. | 15:45:57 |
|  dantefromhell joined the room. | 17:04:57 |
| dantefromhell | I'm looking for a code example to make `extraConfig` pass a `secret.key` file into `vmWithDisko`.
So I can test my encrypted ZFS partitioning scheme | 17:10:40 |
| dantefromhell | @brian:bmcgee.ie have you been able to solve this? | 18:16:56 |
 @brian:bmcgee.ie | Afraid not | 18:19:26 |
| dantefromhell | @joerg:thalheim.io
as of commit 6c1b834 produces a syntax error during each
```
++ dirname /nix/store/4vgl19l2swbfvq2x8d7qrsf8qchrypbj-disk-deactivate/disk-deactivate
+ jq -r --arg disk_to_clear /dev/nvme0n1 -f /nix/store/4vgl19l2swbfvq2x8d7qrsf8qchrypbj-disk-deactivate/disk-deactivate.jq
jq: error: syntax error, unexpected ']' (Unix shell quoting issues?) at , line 49:
]
jq: 1 compile error
```
commit 8d6dd03 works correctly. | 19:59:40 |
| dantefromhell | * @joerg:thalheim.io
as of commit 6c1b834 produces a syntax error during each
++ dirname /nix/store/4vgl19l2swbfvq2x8d7qrsf8qchrypbj-disk-deactivate/disk-deactivate
+ jq -r --arg disk_to_clear /dev/nvme0n1 -f /nix/store/4vgl19l2swbfvq2x8d7qrsf8qchrypbj-disk-deactivate/disk-deactivate.jq
jq: error: syntax error, unexpected ']' (Unix shell quoting issues?) at <top-level>, line 49:
]
jq: 1 compile error
commit 8d6dd03 works correctly. | 19:59:51 |
| dantefromhell | havent been able to spot the syntax error | 20:00:48 |
| food style edible product | Anyone have a btrfs/disko/impermanence config I can peep? Having trouble getting conflicting definitions for filesystems."/foobar" or having stage1 failures due to mounts not working. | 21:47:38 |
| food style edible product changed their display name from Department Store Crust-Punk-Chic to FORWARD TO EXTINCTION. | 21:54:13 |
| food style edible product changed their display name from FORWARD TO EXTINCTION to Like/Share/Subscribe. | 21:54:53 |
| 12 Mar 2025 |
|  cornfused joined the room. | 01:00:21 |
 Mic92 | @dantefromhell:matrix.org: is this where I added the change for swap? | 04:25:45 |
| food style edible product changed their display name from Like/Share/Subscribe to KFC is All I See. | 05:25:10 |
| dantefromhell | In reply to @joerg:thalheim.io
@dantefromhell:matrix.org: is this where I added the change for swap? yes. sorry for the mangled message, editor on the phone displays the code highlights properly, but f*** it up once I hit save/ send | 12:52:57 |
| 13 Mar 2025 |
 waltmck | I have a ZFS mirror running as root that was formatted with disko. If I replaced both drives one at a time (i.e. with zpool offline and zpool replace), would the initramfs know where to look to mount the root partition next time I boot? | 15:58:40 |
| waltmck | in my disko config the drives are referred to as /dev/disk/by-id identifies, so I want to make sure that there is no hard-coded fstab entry or anything that uses those | 15:59:56 |
| waltmck | * in my disko config the drives are referred to as /dev/disk/by-id identifiers, so I want to make sure that there is no hard-coded fstab entry or anything that uses those | 16:00:27 |
| food style edible product changed their display name from KFC is All I See to That's "Mr. Anarchist" to you. | 17:48:08 |
| 14 Mar 2025 |
 phaer | In a typical zfs setup, fstab doesn't reference the underlying devices but only the datasets (i.e. rpool/root, not /dev/disk/by-id/...) here.
the initrd pretty much does a zpool import on boot, so you should be good after zpool replace. (but no warranty and all ;)) | 15:46:25 |
| phaer | * In a typical zfs setup, fstab doesn't reference the underlying devices but only the datasets (i.e. rpool/root, not /dev/disk/by-id/...) here.
the initrd pretty much does a zpool import on boot, so you should be good after zpool replace. (but no warranty as usual ;)) | 15:46:32 |
| waltmck | I see. So I understand that it works roughly like: ZFS scans all storage devices for zfs disks, and then makes the pools it finds available which can be referred to by name in the legacy mount scripts? | 15:48:01 |
| phaer | Yes, exactly | 15:48:34 |
