!RROtHmAaQIkiJzJZZE:nixos.org

NixOS Infrastructure

352 Members
Next Infra call: 2024-07-11, 18:00 CEST (UTC+2) | Infra operational issues backlog: https://github.com/orgs/NixOS/projects/52 | See #infra-alerts:nixos.org for real time alerts from Prometheus.109 Servers

Load older messages


SenderMessageTime
26 Jun 2025
@ctheune:matrix.flyingcircus.ioChristian Theuneeta 25 minutes20:14:40
@raitobezarius:matrix.orgraitobezarius
In reply to @qyliss:fairydust.space
raitobezariuscould presumably also forward it to you as an attachment
I have the mbox on another server right now :<
20:15:00
@ctheune:matrix.flyingcircus.ioChristian Theuneno worries, fill me in when you can. if lore comes back i should be able to get the message myself.20:15:57
@ctheune:matrix.flyingcircus.ioChristian Theuneyup.20:38:19
@ctheune:matrix.flyingcircus.ioChristian Theunemy test is clean now, too, with 6.12.34, and the revert and workaround you mentioned.20:38:35
@ctheune:matrix.flyingcircus.ioChristian Theuneah, lore is back20:41:08
@ctheune:matrix.flyingcircus.ioChristian Theunekinda ... 20:42:03
@ctheune:matrix.flyingcircus.ioChristian Theuneguess i was just lucky with that 1 request. 20:42:14
@ctheune:matrix.flyingcircus.ioChristian Theunewell, off to bed now for me. i'll try to respond tomorrow in the mailing list thread.20:43:24
@raitobezarius:matrix.orgraitobezarius
In reply to @ctheune:matrix.flyingcircus.io
my test is clean now, too, with 6.12.34, and the revert and workaround you mentioned.
Happy to hear that my root cause analysis seems valid!
21:19:05
@sofie:fsfe.org@sofie:fsfe.org left the room.22:16:52
27 Jun 2025
@ctheune:matrix.flyingcircus.ioChristian Theunelore is back and i'm writing a follow up mail now05:30:32
@zhaofeng:zhaofeng.liZhaofeng LiOh wow, thanks for bisecting it. Is there a canonical Lix-side discussion that I can link to?06:22:07
@ctheune:matrix.flyingcircus.ioChristian Theune raitobezarius: i'm trying out the patch combination that dominique suggested and will give him feedback in the next hour. but i'll stress that I have no semantically valid analysis of those patches and am just being a monkey, so I have no idea what makes more sense regarding the logic of the changes themselves. 07:52:52
@raitobezarius:matrix.orgraitobezariusHere: https://matrix.to/#/!lymvtcwDJ7ZA9Npq:lix.systems/$wLqRlm7-iNmrkN2Tcn--Tmi92id4wgvKC5APwiEYYgw?via=lix.systems&via=matrix.org&via=catgirl.cloud but it's mostly me melting down10:05:47
@raitobezarius:matrix.orgraitobezariusYeah, I will jump in when I have some time, I'm waiting for David to chime in because they probably have a better understanding, if nothing happens, I may try my hand at fixing it but this will take some time10:06:35
@raitobezarius:matrix.orgraitobezariusin our case, we just pinned 6.6 for now and called it a day10:06:49
@ctheune:matrix.flyingcircus.ioChristian Theuneyeah, we're running our affected tests without v9fs because we're currently upgrading the fleet to 6.12. we've been running with 6.12 in staging for a while now (but the tests were pinned on 6.6 due to combinatorial explosion) and are investing another potential kernel issue that seems to go away with 6.12 ... 10:07:54
@ctheune:matrix.flyingcircus.ioChristian Theunethere are some days where it feels like luck getting a kernel version that works straightforward "everywhere" ... 10:08:17
@ctheune:matrix.flyingcircus.ioChristian Theunethen again, i'm happy to not be stuck on a distro kernel - this makes talking to upstream much more pleasant10:08:33
@ctheune:matrix.flyingcircus.ioChristian Theuneat some point i need to find a kernel developer, get him a beer and figure out why folios seem to be something that keeps popping up in bugs ... 10:09:06
@raitobezarius:matrix.orgraitobezariusfolios is one of the most complicated technical debt payment ongoing in the kernel10:10:58
@raitobezarius:matrix.orgraitobezariusso it's very normal that they're exploding in many ways given the lack of fstests that can simulate realistic workloads and the classical CI discussion when it comes to the kernel10:12:07
@emilazy:matrix.orgemilylearning that Linux filesystems don't get deeply rigorous automated testing for every release is like nerd Santa10:13:19
@uep:matrix.orguepI mean, quite honestly, that's one of any number of reasons to use zfs10:16:51
@uep:matrix.orguepWith all the obvious irony that carries10:17:20
@raitobezarius:matrix.orgraitobezarius (plot twist: ZFS don't get deeply rigorous automated testing for every release neither :P — though they have a generous manual test suite — ZTS — which is run on every staging release manually by ZFS leadership) 10:18:10
@raitobezarius:matrix.orgraitobezariusas always, I recommend: https://lwn.net/Articles/1015320/ to learn more about folios (a lot of good backlinks)10:19:18
@uep:matrix.orguepI know, zfs releases rather than every kernel. That's part of the irony10:20:35
@raitobezarius:matrix.orgraitobezariusSpeaking of the wolf, David answered10:26:48

Show newer messages


Back to Room ListRoom Version: 6