Colmena | 330 Members | |
| A simple, stateless NixOS deployment tool - https://github.com/zhaofengli/colmena | 115 Servers |
| Sender | Message | Time |
|---|---|---|
| 21 Oct 2022 | ||
| in fact, I just built a colmena-compat module for my hive module system | 08:09:20 | |
| nice | 08:10:26 | |
| I should probably throw these things in a repo somewhere, but eeeeh | 08:10:55 | |
| more cleanly separating the two halves of colmena might be a thing in itself | 08:13:42 | |
| * more cleanly separating the two halves of colmena might be a worthwhile thing to do | 08:13:52 | |
| There is more than two halves to it actually | 08:14:00 | |
| this has improved a lot already recently | 08:14:40 | |
| care to explain? | 08:15:07 | |
| there are several representations that get produced along the way:
now all these transitions between the representation are currently quite entangled into one big nix function that goes all the way from your colmena hive to colmenaEval what has improved lately is, that you can decide to call this makeHive function yourself and provide colmena a pre-processed colmenaEval attrset this also makes it possible to write other functions that produce the same format, but from a different source format | 08:27:04 | |
| and now we can write other useful functions to transform between these representations:
and so on | 08:36:59 | |
| * there are several representations that get produced along the way:
now all these transitions between the representation are currently quite entangled into one big nix function that goes all the way from your colmena hive to colmenaEval what has improved lately is, that you can decide to call this makeHive function yourself and provide colmena a pre-processed colmenaEval attrset this also makes it possible to write other functions that produce the same format, but from a different source format | 08:43:42 | |
oh yes, makeHive is a major improvement already | 11:31:15 | |
| Yes, but colmenaEval is really just nixosConfigurations in a trench coat, and so conversion functions from/to nixosConfigurations could be useful | 11:52:20 | |
| so there is still room for improvement | 11:52:27 | |
| I keep wanting a multi-host service abstraction | 11:52:57 | |
| we have a few of these strung about in an ad-hoc manner in this codebase | 11:53:07 | |
| but they're all bad in different ways | 11:53:14 | |
| you can already kinda do that with hive by accessing the nodes specialArgs attribute | 11:53:19 | |
| yes, we have that | 11:53:28 | |
| and some ad-hoc conventions on top | 11:53:33 | |
| sometimes we filter nodes by name, other times by tag etc. | 11:53:55 | |
| I also generate a
| 11:54:09 | |
| o nice | 11:54:17 | |
| * I also generate a
| 11:55:28 | |
| * I also generate a
| 11:55:40 | |
| * I also generate a
| 11:55:52 | |
| makes some of the modules way cleaner | 11:56:02 | |
| this just makes one group per tag, right? | 11:56:32 | |
| yes | 11:56:43 | |
like the wireguard module just adds deployment.tags = [ "wireguard" ]; and then accesses groups.wireguard to get all the hosts that have the module enabled (it should mesh to) | 11:57:13 | |