<samueldr>
worldofpeace: are they technically valid?
<samueldr>
those that are valid and not open-ended, and those that are actionable I think should be kept, no?
<worldofpeace>
samueldr: Well there's a lot of package requests that actually have pr resolving them. I'll help merge those. Some of them I'm not sure
<gchristensen>
the anser is, I think, treat them nomrall
<gchristensen>
normally
<samueldr>
right
<worldofpeace>
Yep will do that :)
<worldofpeace>
I'll just mentally separate them from the reporter with that techincal focus
<gchristensen>
sounds perfect
<ekleog>
any idea why hydra's setting to send mails to maintainers when a build fails has never been turned on again after it was turned off? (cc gchristensen as someone with full hydra access)
<gchristensen>
no idea, I wonder if it has to do with "nrNotificationsPending" : 628450,
<samueldr>
wasn't it because a failure in something like stdenv sent mail to *everyone*?
<gchristensen>
we created a new channel jobset and then a lot of stuff failed "for the first time"
<samueldr>
on march 11th
<ekleog>
maybe it'd make sense to drop all pending notifications and turn it on again? was kind of helpful
<gchristensen>
it was
<gchristensen>
maybe open n issue on nixos-org-configurations?
pie___ has quit [Remote host closed the connection]
Lisanna has joined #nixos-dev
Lisanna has quit [Client Quit]
<infinisil>
tilpner: Okay c is now blacklisted
<tilpner>
c++ infinisil++
<{^_^}>
infinisil's karma got increased to 46
<infinisil>
Karma strategy: break bot, then fix it, profit :P
<Synthetica>
Gotta get graham's stickers
Jackneill has quit [Ping timeout: 260 seconds]
<init_6>
I have a few fun patches for all linux kernel versions. first it`s updated cko https://www.linuxpages.org/cko_en.php second zen-kernel based - additional kernel logos plus show random available logo. logos can see there https://init-6.bitbucket.io/content/2017/01/Очередное-обновление-bentoo-sources/ and finally patch which adds ISO-Latin-1 font used in apple`s XNU kernel ported with
<gchristensen>
LnL: it doesn't have quite all the same stats yet, but most of 'em yeah :D
<init_6>
I'm trying to add these patches to default nixos kernel
<Synthetica>
init_6: "Kernel logos?" On boot you mean? Maybe we shouldn't be providing art from random other distros and anime girls? (Plus, if I understand correctly, which isn't guaranteed, because it's in Russian(?), the same can already be achieved with `boot.plymouth.logo`)
Jackneill has joined #nixos-dev
<init_6>
Synthetica: "On boot you mean?" yep. all logos from zen-kernel + old kernel logos like Tasmanian devil from 2.6.29
<init_6>
there is a lot of stuff
<gchristensen>
init_6: I'm not sure those patches would be accepted in to nixpkgs
<ekleog>
woooo for you, you're the ones who wrote that RFC :p
<gchristensen>
yeah but you got the harder job :)
<gchristensen>
great picks, globin
<globin>
gchristensen: now only have to incorporate all the comments this afternoon/evening :)
orivej has joined #nixos-dev
orivej has quit [Ping timeout: 240 seconds]
worldofpeace has joined #nixos-dev
<timokau[m]>
Just reiterating that I really think ekleog ought to have commit access, he's one of the more active community members right now (way more active than a lot of people with access) :)
<gchristensen>
domenkozar: ^
<ekleog>
timokau[m]: actually got it a few days ago :°
<gchristensen>
nice
<timokau[m]>
Oh, didn't notice
<timokau[m]>
Congrats then :) Very good to have you
<ekleog>
:)
<ekleog>
(and thank you for the recommendation!)
<ekleog>
gchristensen: fwiw, the mac-guest are not at 100% disk but the mac-host are at 100%disk, and both -guest and -host are at 100% inodes… can't remember whether you already deployed the auto-gc stuff?
<ekleog>
(well, technically stable at 99.98%, I assume it's equivalent to 100%)
<gchristensen>
"free" :)
<ekleog>
oooh
* ekleog
should learn how to read
<ekleog>
:D
<gchristensen>
:D <3
<gchristensen>
ekleog: using zfs makes those appear a bit funny :)
<ekleog>
hmm? I know nothing about zfs
<gchristensen>
well there is no specific inode limit beyond "is there enough space on the disk to write another inode?"
<ekleog>
oh ok :)
<gchristensen>
and also the way the VM is stored (as a ZFS "vdev" device) it doesn't take up space on the / filesystem
<gchristensen>
so from there the only thing on the / is enough of a nixos system to run qemu, which is not very much
<ekleog>
nice :)
<ekleog>
btw, do you know the right prometheus|grafana query to know the number of build jobs queued on darwin? there's a rebuilding since like ~27hrs, it's getting almost weird that it's still not started
<ekleog>
+queued
Profpatsch has quit [Ping timeout: 246 seconds]
* ekleog
waiting for rustc.x86_64-darwin to complete to trigger an ofborg build (a)
<gchristensen>
so... the data in prometheus is data I haven't worked with before. when I imported it in to my personal prom, it was in a different format. let me see ...
<simpson>
What are you trying to learn/graph? That's the question I ask myself when I'm at the Prom console.
<ekleog>
oh so it's just not on status.n.o/prometheus :)
<simpson>
ekleog: It's doing the right thing; you can also check /metrics
<ekleog>
hmm wait, there is no hydra_builds_queued in this list
<simpson>
gchristensen: Looking great! One thing to know about that isn't yet a problem: You have to pay for each *label* used, not each *dimension* declared. Another way to look at it is that scrapes have a CPU cost that is linear in the /metrics size.
<simpson>
So hopefully the number of build hosts, jobsets, and machine types is very slow-growing.
<gchristensen>
simpson: right, jobsets is for sure -- build hosts we might have a several thousand over six months
<gchristensen>
but I think that is desirable information to record
<gchristensen>
simpson: there is _much_ more information I'd like to collect -- you can see the source here: http://hydra.nixos.org/queue-runner-status -- if you have advice, I'd love it ... even better would be a PR :P
<simpson>
Of course! Just a heads-up. IP addresses are one of those thing that are risky to store as labels.
<simpson>
How are you estimating S3 cost, out of curiosity?
<gchristensen>
hydra does it :)
<simpson>
(FWIW my free advice is probably worth more, both in $$ and in usefulness. I hate writing exporters; that's why I like reusing the community code.)
<simpson>
Ah. Probably not a model I can immediately borrow, then. But still worth looking into.
<gchristensen>
that is one of the only "calculated" metrics I re-exported from the json since I'm definitely not going to reimplement that as a prom query. :)
<qyliss^work>
gchristensen: fyi, I just did a multi-user Darwin install in a VM. `nix-channel --list; sudo nix-channel --list` outputs nothing
<qyliss^work>
I think there's definitely something wrong there
<gchristensen>
must be, I can't look in to it now, sorry :/
<qyliss^work>
no worries
<simpson>
Precalculation of stuff like that makes sense, too, since it results in fewer metrics exported and scraped.
<qyliss^work>
Was mostly just looking for confirmation that I was right about something being wrong
<qyliss^work>
Ah, I was wrong anyway
<qyliss^work>
You can't do `sudo nix-channel --update` because it uses HOME
<qyliss^work>
Which is preserved by sudo
<qyliss^work>
That's fairly unintuitive, though... wonder if that's how it should be
<gchristensen>
qyliss^work: I wonder how it "should" be too. maybe the user who installs it is in "charge" of the system, ie has the "root" profile
<gchristensen>
(by "root" I mean the nix daemon comes from this user, however the profile is still named after their user)
<qyliss^work>
I wonder if maybe the user who installs it should just get nixpkgs added to their own profile?
<qyliss^work>
You end up with two profiles, but I think that's what I would expect from a multi-user install - one for root and one for the user?
<gchristensen>
for linux yes, but for macos not sure
<gchristensen>
the root account files gets blown away periodically, making it not at all a good pick
<qyliss^work>
I don't think the installer should make the user who installed "special"
<qyliss^work>
What is root's profile even used for?
<qyliss^work>
(I don't really use profiles at all, but am motivated to try to get them to work for people coming to Nix)
<gchristensen>
the nix daemon comes from it, and the profile script referenced in /etc/bashrc, and also the default NIX_PATH
<qyliss^work>
The default NIX_PATH doesn't include a user's own channels ,right?
<LnL>
yeah only nix-env finds those because it has weird behaviour and doesn't honor NIX_PATH
<qyliss^work>
Should it?
<qyliss^work>
It seems very strange that a multi-user install is actually more of a "root" install, and every user uses root's stuff.
<gchristensen>
probably the thing to do is describe what it should do, see if people agree, and then implement :)
<qyliss^work>
Maybe I'll write a Discourse post
<gchristensen>
yes please!
<qyliss^work>
I almost said "RFC" until I remembered that RFC about how RFCs don't get accepted :P
<gchristensen>
let's see how the new RFC edits happen
__Sander__ has quit [Quit: Konversation terminated!]
<globin>
WRT the RFC improvement RFC, I'm inclined to add the following, any comments?
<globin>
" The author and co-author cannot be part of the Shepherd
<globin>
Team, half of the Shepherd Team or less can also be part of the RFC Steering
<ekleog>
Same here, the idea is good but the wording weird, maybe something like “The author and co-author cannot be part of the Shepherd Team. In addition, at most half of the Shepherd Team can be part of the RFC Steering Committee”?
<globin>
sounds even better, thank you!
<ekleog>
:)
<Ericson2314>
+1 new wording too
<gchristensen>
globin: "author nor coauthor"?
<gchristensen>
"neither the author or coauther can be part of ..."
<gchristensen>
other part seems good!
<globin>
might become obsolete if I can decide to ditch the co-author
<shlevy>
Do we want to add anything about an author (or coauthor, if we keep those) recusing themselves from steering committee decisions relating to their PR?
<shlevy>
(haven't seen the thread here, just saw the commit pushed)
<globin>
hmm, since nomination by steering committee has to be unanimous I don't think it's necessaryy
<shlevy>
Fair
init_6 has quit [Ping timeout: 250 seconds]
phreedom_ has joined #nixos-dev
phreedom has quit [Ping timeout: 256 seconds]
hedning has joined #nixos-dev
pie__ has quit [Remote host closed the connection]
<{^_^}>
rfcs#36 (by globin, 1 week ago, open): Improving the RFC process
orivej has joined #nixos-dev
<thoughtpolice>
gchristensen: Just want to thank you for ofborg, just helped me quickly find and fix a Darwin failure (I hope!)
<gchristensen>
yay!
<gchristensen>
LnL will be very delighted to hear that :)
<LnL>
I'm not sure how you did that with the current state of master, but yay :D
<LnL>
gchristensen: btw, sometime around nixcon somebody opened up a wip pr and pushed about 20 commits to fix a darwin build!
<gchristensen>
:o :D
<LnL>
not sure how I should feel about it, but that's some dedication
<gchristensen>
I've been thinking about this, if your vm can be made disposable (daily?) and also protect your internal network, it'd be nice to open up darwin to everyone
<LnL>
I use the machine to build other stuff from time to time, but I'm down for that if the load is reasonable
<LnL>
for cases where I want the compute to test something I already reset it and disable the daemon for a bit
<gchristensen>
nice
<gchristensen>
:/ I can't find my mac on my network.
<LnL>
heh
<LnL>
ideally the credentials would even be rotated so it doesn't contain any long living secrets