| 27 Jul 2025 |
 emily | I thought by job slot you meant like, CLs | 16:16:10 |
| emily | not max-jobs | 16:16:12 |
 raitobezarius | sorry | 16:16:16 |
| raitobezarius | job slot in my jargon refers to buildkite slots | 16:16:23 |
| raitobezarius | 1 slot is consumed by one derivation in our CI currently | 16:16:33 |
| emily | I know nothing of Buildkite :) | 16:16:35 |
| emily | ok | 16:16:37 |
| emily | per derivation or per leaf | 16:16:45 |
| raitobezarius | per derivation | 16:16:50 |
| raitobezarius | uhm | 16:16:53 |
| raitobezarius | per leaf | 16:16:53 |
| raitobezarius | sorry | 16:16:54 |
| emily | like what is the Nix-side concurrency | 16:16:56 |
| raitobezarius | per leaf derivation | 16:16:58 |
| raitobezarius | In reply to @emilazy:matrix.org
like what is the Nix-side concurrency you mean in terms of max-jobs? | 16:17:15 |
| emily | and cores yes | 16:17:20 |
| raitobezarius | something like 4C | 16:17:39 |
| raitobezarius | 2-3 max jobs probably | 16:17:44 |
| emily | ok | 16:24:27 |
| emily | do you know the specs of the M2s? | 16:24:30 |
| raitobezarius | the current CI is a M1 | 16:24:45 |
| raitobezarius | the M2 I can give it to you, give me a sec | 16:24:49 |
| raitobezarius | https://www.ebay.com/itm/146539190930 | 16:25:15 |
| emily | 8 GiB :'( | 16:28:20 |
| emily | two VMs probably not viable then | 16:28:48 |
| emily | if you can get Buildkite to only assign slots from the same CL then it seems fine to do 1 | 16:29:24 |
| emily | you need some kind of mutex-y or wrapper thing anyway because you need to make the wipe happen between | 16:29:42 |
| emily | tbh if you have enough Macs then you might be able to skip the VMs and just do MDM wipes between builds and rotate through the machines | 16:30:05 |
| emily | not sure if the total throughput looks better or worse like that | 16:30:10 |
| emily | I do trust MDM wipes on Apple Silicon to be safe from all but the most sophisticated userspace attacks and I don't trust the latter to consider escaping a VM an obstacle | 16:31:13 |
