| 19 Feb 2025 |
 @adam:robins.wtf | how does the coldsnap path help? | 22:39:34 |
 Arian | coldsnap bypasses the vmimport service | 22:39:43 |
| Arian | which god knows what it’s written in | 22:39:47 |
| Arian | it just uploads blocks directly to EBS | 22:39:53 |
| Arian | image uploads take like 5 seconds with coldsnap | 22:40:27 |
| @adam:robins.wtf | that sounds better than the current path. the timeout it's hitting on import is 10 minutes | 22:40:52 |
| Arian | what region are you in?t | 22:42:34 |
| @adam:robins.wtf | us-east-2 | 22:43:31 |
| Arian | that timeout is set by AWS. usually AWS timeouts are pretty aligned with the normal behaviour of services | 22:43:37 |
| Arian | might be AWS is having a bad time | 22:43:38 |
| @adam:robins.wtf | the imports are completing, but only after the waiter times out :/ | 22:50:10 |
| 20 Feb 2025 |
|  @hexa:lossy.network left the room. | 02:39:47 |
| @adam:robins.wtf | It took me longer to upload due to my internet connection, but once the snapshot is imported registering it as an AMI is almost instant. | 13:51:42 |
| @adam:robins.wtf | This route also seems to be more space efficient in Amazon's reading of the size. A snapshot using the current import-snapshot shows as the full size of the disk image used during build, but a coldsnap uploaded image shows the actual disk usage of the image. | 13:53:22 |
| @adam:robins.wtf | * It took me longer to upload due to my internet connection, but once the snapshot is uploaded, then registering it as an AMI is almost instant. | 13:54:27 |
| @adam:robins.wtf | hmm, but that snapshot doesn't seem to work for launching instances. likely i did something wrong :/ | 15:01:49 |
| @adam:robins.wtf | I'm extending the timeout for our use, since a snapshot is still imported even if the waiter times out. But re-running upload-ami will initiate another import-snapshot, ignoring the already imported one | 15:25:30 |
| 22 Feb 2025 |
| Arian | Oh you need to coldsnap over a raw image | 11:01:35 |
| Arian | You're probably coldsnapping the sparse VHD lol | 11:01:49 |
| Arian | Which will not work | 11:01:52 |
| Arian | That's why it shows up as smaller | 11:02:12 |
| Arian | So build the image with format = raw | 11:02:38 |
| Arian | And then upload | 11:02:40 |
| Arian | * That's why it shows up as smaller. You cant upload vhd format. You need to upload the raw block devices bytes. | 11:03:30 |
| @adam:robins.wtf | That makes complete sense. Thanks | 12:59:51 |
| 24 Feb 2025 |
 commiterate | Arian did you get around to trying the fluent-bit PR? | 17:30:44 |
| Arian | It's been on my backlog. Work was really packed the past few weeks. But I'll get to it as we need the changes internally | 19:12:57 |
| 27 Feb 2025 |
|  drewhaven joined the room. | 20:06:37 |
| drewhaven | Redacted or Malformed Event | 20:10:30 |
| drewhaven | Does SSM work will with on-prem NixOS installs? I'm looking to set up a bunch of headless NUCs that are deployed to a bunch of different locations. I want the predictability of Nix configs and flakes, but I'm not sure how the remote management will work. SSM seems to imply that it wants a mutable system to manage, but I guess that's just handled with some scripts that use Nix tools for changes, upgrades and rollbacks? | 20:21:51 |
