| 3 Jun 2021 |
 Arian | Has anybody worked on pull-based deployment models vs push-based? | 11:06:39 |
| Arian | e.g. I want my nodes to query some metadata server periodically and then activate the new config; instead of requiring SSH access to push the configs | 11:07:02 |
| Arian | it sounds conceptually easier to set up, esp in things deployed in edge environments or coorporate environments where machines are often not addressable from the outside | 11:07:42 |
| Arian | while true; deriv=$(curl https://metadata.local/derivation); nix-store --realise $deriv; $deriv/switch-to-configuration switch; done | 11:08:46 |
 andi- | Arian have you looked at the hail module in nixos? | 11:40:21 |
| Arian | Nope. TIL! | 12:33:36 |
 Linux Hackerman | I only recently realised that it's named for "hail hydra" | 12:37:17 |
 Jez (he/him) | 🐙 | 12:41:37 |
 Buckley | In reply to @linus.heckemann:matrix.mayflower.de
Buckley: it's not really elegant, but to at least see if the nixpkgs they were built from is up-to-date I usually run `nixops ssh-for-each --parallel nixos-version` hmmm 🤔 ok. maybe ill go back to morph and see how easy it would be to work this into a custom health check | 13:44:17 |
|  ajz229 joined the room. | 22:57:29 |
| ajz229 set a profile picture. | 22:59:44 |
| 4 Jun 2021 |
| Jez (he/him) left the room. | 10:44:44 |
 manveru | Buckley: another option would be to compare the link /run/current-system with whatever you want to deploy | 20:03:14 |
| Buckley | yeah, a manual script that grabs the system paths from morph and checks each system seems like the best option currently | 20:04:47 |
| Buckley | i've sort of settled on morph. it most closely matches my goals. although its unfortunate they dont natively support flakes yet | 20:05:51 |
| manveru | yeah... i still mostly use good old nixos-rebuild ^^; | 20:06:28 |
| Buckley | lol | 20:06:36 |
| Buckley | i got to a place where i kind of wanted tagging, and thats just one feature too much for my bash scripts | 20:07:07 |
| manveru | ported some of my machines to deploy-rs, and it's also nice, and fully compatible with nixos-rebuild if used with flakes | 20:07:32 |
| Buckley | deploy-rs was actually the first one i tried, but its super feature-bare. like it doesnt have an exec function as far as i can tell | 20:08:17 |
| Buckley | to be fair it does exactly one thing pretty well | 20:08:32 |
| manveru | that's what ssh is for... | 20:08:40 |
| manveru | indeed :) | 20:08:48 |
| manveru | it'd be super simple to write an exec on top of it though | 20:09:10 |
| manveru | ssh nix eval --raw .#deploy.nodes.abaddon.hostname -- echo hi | 20:09:53 |
| manveru | like an alias for that | 20:09:56 |
| Buckley | i've been slowly building nixos machines at work, so far nobody has told me to stop it lol, but i'd like to load as much of the work as possible into a tool, so that when people ask how it works i dont hand them a hive of bash scripts | 20:10:04 |
| manveru | uh, matrix doesn't like nesting those quotes | 20:10:15 |
| Buckley | outer
`inner`
| 20:10:34 |
| Buckley | (just testing what that did ^) | 20:10:45 |
