| 21 Sep 2025 |
 hexa | anyway, these are my thoughts. enjoy | 21:30:58 |
 dgrig | ECC would be good to have, but I wonder if those intel 13900s are affected by this or did Hetzner RMA'd them/patched them all fast enough | 21:35:57 |
| hexa | supposedly fixed by firmware updates a while ago | 21:37:17 |
|  Nico joined the room. | 21:42:45 |
 Tom | I think you'll end up paying and receiving (if you don't tell them to remove those) the default two 1tb drives anyway. So if you add 2x 1tb you're getting 1 TB, 1 TB, 2 TB and 2 TB. Just mentioning it in case that wasn't considered and somehow affects the plans. | 21:45:14 |
| hexa | yeah, I'm aware. | 21:45:53 |
| hexa |
ECC would be good to have
ECC has been doing its thing on haumea for a few months, so we're not going to give that up
| 22:33:24 |
| hexa | the primary reason we're making a move | 22:33:47 |
| hexa | MCE records summary:
9168 Corrected error, no action required. errors
| 22:34:00 |
| hexa | * Memory controller events summary:
Corrected on DIMM Label(s): 'mc#0csrow#1channel#0' location: 0:1:0:-1 errors: 4973
MCE records summary:
9168 Corrected error, no action required. errors
| 22:34:20 |
| 22 Sep 2025 |
|  @felix.schroeter:scs.ems.host changed their display name from Felix Schröter to Felix Schröter (🌄 29.09. – 05.10.). | 09:55:59 |
 Vladimír Čunát | Some Hydra metrics are gone now:
https://grafana.nixos.org/d/MJw9PcAiz/hydra-jobs | 16:09:52 |
| Vladimír Čunát | I think I saw them working less than 24h ago. | 16:11:08 |
 ghpzin | instance label changed, so dashboard shows nothingjm | 16:18:44 |
| ghpzin | instance label changed, so dashboard shows nothing, metrics seem to be still there | 16:19:05 |
| ghpzin | instance label changed, so dashboard shows nothing:
https://grafana.nixos.org/goto/DdQG9uCNg?orgId=1 | 16:25:43 |
| ghpzin | instance label changed from monitoring.nixos.org:9200 to localhost:9200, so dashboard shows nothing:
https://grafana.nixos.org/goto/DdQG9uCNg?orgId=1
https://github.com/NixOS/infra/commit/0ebab0affe655b6a278843124a10637c50e6e78d#diff-7dbeda5809ed539867d6bc0ec13abd3ecfd987097db6c868533588c2e36b3d82L54 | 16:48:41 |
| hexa | yeah, was fairly pointless to expose the metric externally | 17:08:37 |
| hexa | unclear why grafana shows no result, but with prometheus I see the metric | 17:12:13 |
| hexa | https://prometheus.nixos.org/query?g0.expr=hydra_steps_queued&g0.show_tree=0&g0.tab=table&g0.range_input=1h&g0.res_type=auto&g0.res_density=medium&g0.display_mode=lines&g0.show_exemplars=0 | 17:12:20 |
| ghpzin | Metrics are still there, dashboard hardcodes instance="monitoring.nixos.org:9200" | 17:15:55 |
| Vladimír Čunát | I think I fixed the graphs now. | 17:16:37 |
| Vladimír Čunát | Just edited the dashboard. | 17:17:02 |
 sterni (he/him) | Do any (Darwin) builders use Nix 2.30 already? We've noticed the interesting problem that the new build-dir location causes longer paths which can relatively easily cause failures becasue unix socket paths are quite restricted on macOS apparently. (Observed with `haskellPackages.network.) | 19:50:07 |
 emily | IIRC this was mitigated in the latest patch release | 19:50:54 |
| emily | with possible further changes to come | 19:51:07 |
| emily | (the limits are similarly strict on Linux btw, just the sandbox helps there) | 19:51:21 |
| hexa | no, they track nixos-stable | 19:51:33 |
| hexa | * no, they track nixpkgs-stable | 19:51:36 |
| sterni (he/him) | https://github.com/NixOS/nix/pull/13839? | 19:51:40 |
