| 9 Dec 2025 |
|  adamcstephens changed their profile picture. | 17:48:34 |
| 10 Dec 2025 |
| adamcstephens changed their profile picture. | 14:49:57 |
| 8 Aug 2025 |
|  0x4A6F joined the room. | 06:32:10 |
| 9 Aug 2025 |
 commiterate | In case someone here has tried: https://discourse.nixos.org/t/nixos-disk-images-with-ephemeral-root-volume-cache-overlay/67821 | 03:03:23 |
| 20 Aug 2025 |
 Arian | Have you looked at the repart appliance image tests? | 11:31:36 |
| Arian | What I'd do is make the AMI EBS Volume an EROFS volume. Then overlay mount the rootfs over it | 11:33:05 |
| Arian | and that can be on the instance store or whatever | 11:33:30 |
| Arian | or just... only mount /var to the instance store | 11:33:37 |
| Arian | Do you want to be able to have interactive nix usage afterwards? | 11:33:44 |
| 21 Aug 2025 |
| commiterate | These tests?
https://github.com/NixOS/nixpkgs/blob/02489e616d84df58c35f0bef559377f454abf410/nixos/tests/appliance-repart-image.nix#L28 | 02:43:36 |
| commiterate | I considered overlayfs but it doesn't seem to cache reads from the lower filesystem (whatever's on the EBS volume) to the upper filesystem (whatever's on the instance store). | 02:45:40 |
| commiterate | I'm not too sure what counts as interactive Nix usage, but these will be CI/CD runners so users should be able to do a nixos-rebuild switch afterwards to modify the environment (e.g. set up Docker Engine). | 02:47:12 |
| commiterate | That makes mounting something at / attractive since users don't need to remember to write under a special directory to hit the instance store instead of the root EBS volume. | 02:58:36 |
| commiterate | * That makes mounting something transparent at / attractive since users don't need to remember to write under a special directory to hit the instance store instead of the root EBS volume. | 03:02:34 |
| Arian | Ah so you're worried that all the nix store paths in the lower store will have slow access? | 12:26:42 |
| commiterate | Yup. | 16:25:30 |
| 28 Aug 2025 |
| Arian | @commiterate:matrix.org: we're getting to something very similar to what you want internally at Mercury for our runner fleet | 11:50:52 |
| Arian | Ill share when I'm at the office | 11:50:57 |
| Arian | Redacted or Malformed Event | 11:51:40 |
| Arian | Redacted or Malformed Event | 11:52:02 |
| Arian | Redacted or Malformed Event | 11:52:29 |
| Arian | Redacted or Malformed Event | 11:52:41 |
| 29 Aug 2025 |
|  @luna-null:matrix.org left the room. | 02:40:24 |
| 31 Aug 2025 |
|  Edgar Lee joined the room. | 09:21:04 |
| 8 Sep 2025 |
|  Inayet set a profile picture. | 02:15:57 |
| 14 Sep 2025 |
|  Emma [it/its] joined the room. | 08:41:04 |
| 23 Sep 2025 |
|  kenji changed their display name from a-kenji to kenji. | 10:42:38 |
| 29 Sep 2025 |
| adamcstephens joined the room. | 14:56:38 |
| adamcstephens | Why are fetch-ec2-metadata.service and apply-ec2-data.service separate services? | 14:57:10 |
| adamcstephens | I'm troubleshooting one or more logic bugs in the fetch script, which will then exit successfully even if it fails, but since apply depends on fetch it makes me wonder why they're separate | 14:59:07 |
