| 25 May 2023 |
 raitobezarius | Basically they were running a giant NFS stuff with a lot of uid for all the users | 17:41:04 |
| raitobezarius | And it collided with Nix | 17:41:07 |
| raitobezarius | So Bad Experience™ | 17:41:13 |
 graham | I'd be happy to hear more from them :) | 17:41:27 |
| raitobezarius | They are probably busy right now but I can try to get them to retry your script Graham! :) | 17:41:40 |
| graham | well, not yet! they won't get 0.9.0 probably | 17:41:51 |
| raitobezarius | Alright, will wait for 0.9.0 and proceed to nerdsnipe them then :) | 17:42:03 |
 Zhaofeng Li | In reply to @raitobezarius:matrix.org
Basically they were running a giant NFS stuff with a lot of uid for all the users Ah ok, I thought it was about having a global nix store shared among all nodes (it would be quite cursed I think) | 17:42:32 |
| raitobezarius | They didn't adopt Nix "far" enough for that yet | 17:42:55 |
| Zhaofeng Li | The school cluster I was talking about has NSS/LDAP auth plus NFS automounts all over the place for home directories and software | 17:44:02 |
| Zhaofeng Li | Typical HPC setup I think, and they wanted to apply the same if they agreed to Nix | 17:44:37 |
| Zhaofeng Li | They are using this for software (PTSD warning if you've seen this stuff): https://sourceforge.net/projects/modules/ | 17:46:45 |
| Zhaofeng Li | Apparently I still have access:
$ module avail python
------------ /pkg/modules/modulefiles ------------
python/2.7.10-tf python/3.10 python/3.6.1
python/2.7.11 python/3.10.0 python/3.6.7
python/2.7.12 python/3.2.1 python/3.6.8
python/2.7.14 python/3.3.3 python/3.7.1
python/2.7.2 python/3.3.5 python/3.7.2
python/2.7.4 python/3.5.1 python/3.7.4(default)
python/2.7.8 python/3.5.2
$ mount | grep python
[redacted] on /pkg/python type nfs (ro,relatime,vers=3,rsize=32768,...)
The end-user UX is actually quite good: module add python/3.3.5 and suddenly you have Python 3.3.5 in your PATH
| 17:54:46 |
| Zhaofeng Li | And don't worry, all the software packages are hand-compiled (./configure && make && make install), no GMO | 17:55:40 |
| 27 May 2023 |
|  stroem - sourcegraph joined the room. | 15:17:52 |
|  NixOS Moderation Botchanged room power levels. | 16:40:27 |
| 29 May 2023 |
 K900 | Ana: I'm on here btw if you want something faster than throwing emails around | 19:01:32 |
| 30 May 2023 |
 hexa | had a friend curl the installer from the command in the readme | 19:50:05 |
| hexa | he's on fedora with selinux and the download still points to 0.8.0, which fails | 19:50:20 |
| hexa | but 0.9.1 is out and 0.9.0 is supposed to work with selinux | 19:50:29 |
| hexa | so updating that link would be mighty helpful 🙂 | 19:50:36 |
| graham | we were serving 0.9.0 to 20% of install requests and it was causing some broken configs in in some important cases on github actions, so we rolled it back to 0% | 19:51:15 |
| graham | 0.9.1 is being served to 10% IIRC of install requests now | 19:51:39 |
| graham | they can get exactly 0.9.1 like this:
curl --proto '=https' --tlsv1.2 -sSf -L https://install.determinate.systems/nix/tag/v0.9.1 | sh -s -- install
| 19:52:23 |
| graham | (but ideally they won't memorialize that in any scripts / documentation) | 19:52:40 |
| hexa | alright, thanks for explaining that 🙂 | 19:53:12 |
| graham | yeah :) 0.9.0 also makes the install a lot faster, so we definitely want to get it out there soon | 19:53:34 |
| graham | but I'm scared of breaking stuff all at once and the auto-uid-allocation changes in particular leave me feeling a bit anxious. so, going a bit slow | 19:54:01 |
| graham | and for context, about 95% of our installs are on in CI, and the way we do ramped deploys doesn't have any information that we could use to do smart targeting about if it is in CI or not | 19:56:29 |
| hexa | the install worked just fine | 19:57:58 |
