| 1 Mar 2026 |
|  cbrxyz joined the room. | 06:15:12 |
|  @schrobingus:matrix.org left the room. | 19:54:59 |
| 2 Mar 2026 |
|  neko-hime joined the room. | 20:38:30 |
| 8 Mar 2026 |
|  Matan (ixxie) changed their display name from ixxie (Matan) to Matan (ixxie). | 10:43:00 |
| 9 Mar 2026 |
|  krejgo changed their display name from craige to krejgo. | 05:48:12 |
|  chaiyo joined the room. | 15:55:14 |
| 12 Mar 2026 |
|  sebastian changed their profile picture. | 18:02:38 |
| 13 Mar 2026 |
|  rendakuenthusiast joined the room. | 07:15:09 |
| 14 Mar 2026 |
|  em joined the room. | 04:39:44 |
|  lektrik joined the room. | 21:30:16 |
| 16 Mar 2026 |
|  purepani changed their display name from purepani to purepanid. | 05:55:20 |
| purepani changed their display name from purepanid to purepani. | 05:55:55 |
| 19 Mar 2026 |
|  KeyZox joined the room. | 16:23:15 |
| 20 Mar 2026 |
|  Evrim Ulu joined the room. | 13:03:07 |
| 26 Mar 2026 |
 Yifei Sun | one problem i have with colmena is if the deployment is done over vpn (say tailscale), and the update requires tailscale or nftables to restart, the deployment would fail | 20:57:19 |
| Yifei Sun | maybe we can change the deployment model a little so that instead of running the switch to configuration directly, we start a transient systemd unit with systemd-run so that the activation keeps going even if the network restarts | 20:58:36 |
| 28 Mar 2026 |
 Eli Saado | interesting, I've not noticed this
I know deployrs has this issue too, and they have a feature called magic-rollback, which rolls back to the previous config if it doesn't receive an acknowledgement, which makes it even worse
| 11:21:30 |
| Yifei Sun | i would say this is pretty easy to reproduce
change the deployment config target node ssh host to tailscale address and override that node’s tailscale name or version or smth it should break | 11:23:15 |
| Yifei Sun | i have llm generated slop patch for detached switch to configuration but i dont think it has the quality to be upstreamed 😂 | 11:26:25 |
| Eli Saado | and the new config doesn't get applied? | 11:26:53 |
| Yifei Sun | it errors out | 11:27:10 |
| Yifei Sun | because ssh connection will get terminated and it’s not handled gracefully | 11:27:33 |
| Yifei Sun | imo the best solution to this would be adding a detached bool flag for deployment config, and a -d/—detached cli flag | 11:28:47 |
| Yifei Sun | https://github.com/stepbrobd/colmena/commit/c5b062e5248ebdfeecf5cb3bf96d4302c278bd3d | 11:29:33 |
| Eli Saado | right, and the rest of the activation script never finishes? | 11:31:10 |
| Yifei Sun | i don’t remember the exact behavior but the error log was like ssh 255 and the rest got terminated leaving the system in a intermediate half broken state | 11:32:46 |
| Eli Saado | ah that's bad | 11:33:20 |
| Yifei Sun | i usually use boot goal with reboot flag when this happens and that sometimes hangs 😂 | 11:33:41 |
| Yifei Sun | i figured this is easily solvable by adding tcptimeout and server keep alive time out to smth like 10sec in ssh options but imo it would be better to enable these ssh options by default | 11:34:46 |
| Yifei Sun | if anyone with more rust knowledge would want to collaborate on this i’d love to get this fixed | 11:36:23 |
