| 8 Aug 2023 |
 Zhaofeng Li | Could you confirm that ssh-ing to the host with the control socket also fails to work in this state? | 04:35:38 |
| Zhaofeng Li | There is ServerAliveInterval which may help with stale sockets as well | 04:36:09 |
|  Christina Sørensen joined the room. | 17:18:26 |
| 9 Aug 2023 |
|  Jeff changed their profile picture. | 03:16:11 |
|  @ekeske:wdz.de joined the room. | 08:17:22 |
|  Find me at aleksana:qaq.li changed their profile picture. | 14:42:49 |
| 12 Aug 2023 |
 Winter (she/her) | In reply to @zhaofeng:zhaofeng.li
I've occasionally encountered master sockets getting stuck as well, but it always cleared itself up issue is i can reproduce it consistently, and it only happens when using it via Colmena :/ | 20:19:16 |
| Winter (she/her) | In reply to @zhaofeng:zhaofeng.li
Could you confirm that ssh-ing to the host with the control socket also fails to work in this state? sure, i can try | 20:19:38 |
| @ekeske:wdz.de left the room. | 20:20:26 |
 void | hey Zhaofeng Li any plans for transactional switches like rewinds in the of no confirmation? | 21:37:17 |
| void | just switched on nftable and lost a box. | 21:37:34 |
| void | * just switched on nftables and lost a box. | 21:37:41 |
| void | * just switched on nftables and lost a box. ;( nothing serious but may be helpful for all. | 21:37:57 |
 Wanja Hentze | yeah, a dead man's switch type feature has been requested a bunch | 21:47:30 |
| Wanja Hentze | "magic rollback" | 21:47:36 |
| Wanja Hentze | I'd like it too | 21:47:45 |
| void | I think implementing transactional switching for n boxes would be a nice feature. | 21:50:19 |
| void | maybe the script can be implemented in nix too. | 21:50:42 |
| Wanja Hentze | in my professional opinion distributed transactions don't really work | 21:56:53 |
| Wanja Hentze | but I'm happy to be proven wrong | 21:57:05 |
| Wanja Hentze | you can pull them off with a consensus algorithm but I don't want my deployments to be a CP (in terms of CAP) | 21:58:19 |
| Wanja Hentze | * you can pull them off with a consensus algorithm but I don't want my deployments to be a CP system (in terms of CAP) | 21:58:38 |
 raitobezarius | I don't think the feature should be implemented in the deployment system | 22:31:03 |
| raitobezarius | Magic rollback should be part of NixOS to a certain extent | 22:31:12 |
| raitobezarius | In reply to @whentze:matrix.org
in my professional opinion distributed transactions don't really work well if you care about C or A… | 22:32:09 |
| raitobezarius | hot take: distributed tx works if you prove that you cannot have P in your network to a certain extent | 22:32:39 |
| raitobezarius | checkmate CAP theorem | 22:32:49 |
| Wanja Hentze | If you're going to pretend things never go wrong, might as well not have rollback either | 22:35:44 |
| Wanja Hentze | In reply to @raitobezarius:matrix.org
Magic rollback should be part of NixOS to a certain extent the rollback part yes (and it is), but the "can I still use my deployment tool on this machine" part depends on the deployment tool | 22:38:20 |
| raitobezarius | In reply to @whentze:matrix.org
If you're going to pretend things never go wrong, might as well not have rollback either that's the goal | 22:38:43 |
