Domen Kožar
Domen Kožar
The only generally applicable assumption is that the agent should still be able to connect to the websocket after the deployment is finished. If not, the deployment should rollback. Besides...
If you deploy the wrong script to the wrong OS, you'll get a weird error. But that can be solved by inspecting .drv file and checking the system :+1:
> The rest is great, can I haz it yesterday ser plz? Next week :rocket:
I went for a bit different approach how to specify the rollback script, rollbacks scripts are now keyed by the system: ``` "rollbackChecks": { "x86_64-linux": "", , "aarch64-darwin": "" }...
Deployed! Last bit missing are the docs: https://github.com/cachix/docs.cachix.org/pull/86
Done: https://docs.cachix.org/deploy/reference#deploy-json Will be release as part of Cachix 1.0.0 tomorrow!
I've been thinking about this one as it came up a few times. I'm not sure what's the best way, but I'd prefer it's fixed in NixOS as it's not...
`nixos-rebuild.sh` already builds Nix first, so we can add support to also build nix.conf and link it to correct place and restart nix-daemon.
I plan to tackle this soon.