gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
<elvishjerricco> I'm curious what kind of drive it is
<cole-h> Oh goodie, I had it in my scrollback thanks to tmux
<cole-h> I was wrong
<cole-h> It was 15 cksum errors, 22 read errors, and 8 write errors!
<elvishjerricco> That's... a lot for it to just disappear on the next scrub
<elvishjerricco> Might be a controller or cable going bananas rather than the drive.
<cole-h> Possible. I think this cable I got from my old mobo a few years ago...
<elvishjerricco> Oh wait it still says no known data errors. So it should be all the read and cksum errors were metadata
<elvishjerricco> so it could have repaired them
<elvishjerricco> Still, to immediately get another cksum error is worrying
<cole-h> (:
<elvishjerricco> cole-h: I believe all reparations are logged in... /var/log/syslog? So you can check there
<elvishjerricco> Logged somewhere anyway. Probably visible in dmesg
<cole-h> The only relevant lines from dmesg: https://gist.github.com/cole-h/d78775f3fd6e6ee27bb2fbe6df36e7bb
<elvishjerricco> cole-h: Well since there are 15 instances of the word "zio" in that gist, I'm guessing it's logging something about the cksum errors. Seems to tell you what physical blocks are being repaired.
<elvishjerricco> It's odd to me that so many errors would be metadata errors; but I guess ZFS does try to do things sequentially, so maybe it just wrote that much metadata in a row in one txg
<cole-h> Weirdest part is that the last zio thing happened at 10:11:02, and my last scrub finished at ~13:xx:xx
<cole-h> Well, weird to me that that 1 cksum error didn't show back up in dmesg
<colemickens> just lost a firefox session with hundreds of tabs open. my fault for not believing when people here mentioned having firefox session problems
<colemickens> literally yesterday I was thinking "wow, I've had this session open for weeks without issue" adn then poof, it's just gone.
<colemickens> this emotion is fun.
* cole-h knocks on wood
<cole-h> I haven't lost any Firefox sessions in a long time...
<colemickens> I feel like a sucker. I need something better to keep track of sessions or keep track of "mind map"-y type browsing sessions. This hurts, I had a lot of stuff open that will be hard to find agian.
<cole-h> idk if this will help, but I came across https://addons.mozilla.org/en-US/firefox/addon/simple-tab-groups/ a while ago.
<cole-h> (Probably not, considering it's an addon, but)
<colemickens> I literally just want a text snapshot of the tabs I have open.
<colemickens> In a git repo or something, checkpointed every 4 minutes. So simple, would save me the hours of pain just inflicted.
<colemickens> so many tabs open with stuff I needed to get back to, just .... gone
<cole-h> I know that feeling... sorry :(
<colemickens> I've decided to embrace the "freeing" side of it and am just going to start doing it by hand for now I guess.
<cole-h> A bit more involved and heavy-handed, but maybe setup snapshots for the profile dir?
<colemickens> that's a cool idea
<colemickens> that's big hammer-y, but at least somewhat actionable. thanks for suggesting it.
rajivr has joined #nixos-chat
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-chat
<cole-h> np. I actually do remember accidentally deleting my session recently-ish, and copying an earlier snapshot I made with ZFS saved me.
<cole-h> elvishjerricco: So, I sent the state of my rpool at a certain snapshot, but is there a simple way to send the individual snapshots as well? I could send snapshot by snapshot, but wondering it there's a better way.
<colemickens> I have a weekend coming up dedicated to consolidating backups and figuring out a zfs backup strategy. The fact that I have /home separated without any snapshotting feels embarrassing.
<cole-h> :eyes:
<gchristensen> you can always move it!
<gchristensen> take a snapshot of /, promote the snapshot to a dataset, delete everything but /home/youruser, mv /home/youruser to the root of that dataset, then mount it to /home
<cole-h> :o
<cole-h> That's.... actually really cool wtf
<cole-h> But I'm thinking of switching to zrepl, since it seems to have an easy way to snapshot, send to external disk, then remove snapshot
<cole-h> Giving me snapshots while also keeping free space
<cole-h> elvishjerricco suggested zrepl a while ago, and I've been meaning to research it + make a module for it
<elvishjerricco> cole-h: -R will send all snapshots on a dataset
<elvishjerricco> up to the specified one that is
<elvishjerricco> and yea zrepl looks like everything I need
<cole-h> oh huh
<cole-h> I guess I did already send all my snapshots lol
<cole-h> I forgot zfs list -r rpool doesn't list snapshots unless I add `-t snapshot`
<cole-h> (:
<cole-h> elvishjerricco: That 1 cksum error might have been a fluke? Still got a ways to go, but 67% done and hasn't shown back up.
<elvishjerricco> cole-h: I'd still be worried about that hardware
<cole-h> True. Too bad SSDs are still so expensive :(
<cole-h> I guess it's a good thing that I get my next paycheck on BF lol
<colemickens> I never did find out conclusively if it was bad RAM or a bad NVME/PCI adapter that gave me trouble for a month
<gchristensen> I want to move to zrepl but znapzend is what I have configured and ... uh ... momentum
<lovesegfault> momentum is a nice way to say inertia :P
<gchristensen> yeah that one
<gchristensen> centrifugal force
<gchristensen> (nothing?)
<elvishjerricco> gchristensen: Same. And I have no idea how I'd migrate my existing snapshots for it
<gchristensen> yea
<gchristensen> those tools do not optimise for migration
tilpner_ has joined #nixos-chat
tilpner has quit [Ping timeout: 240 seconds]
tilpner_ is now known as tilpner
<lovesegfault> really cool stuff
<lovesegfault> They've been doing such fascinating work on bootstrapping
<lovesegfault> It really is a shame that Guix is an FSF distro
red[evilred] has quit [Quit: Idle timeout reached: 10800s]
Dotz0cat has quit [Ping timeout: 240 seconds]
Dotz0cat has joined #nixos-chat
pie_ has quit [Ping timeout: 240 seconds]
pie_ has joined #nixos-chat
<V> colemickens: just back up the sessionstore.jsonlz4
<V> optionally, decompress it first
<V> also optionally: pipe it through jq to format it so the diffs are sane (albeit at the cost of increasing the size on disk)
<V> less of an issue if you have a compressing filesystem
<V> (and probably not an issue at all, unless you're like me with 8k tabs open and a 50MiB compressed sessionstore file)
<pie_> lovesegfault:
<pie_> doing gods work
<pie_> possible before, such as:
<pie_> OH MY GOD We’re giving users more flexibility on the command line, with the addition of three package transformation options: --with-debug-info (always debug in good conditions!), --with-c-toolchain, and --without-tests. Transformations are now recorded in the profile and replayed upon guix upgrade. Furthermore, those options now operate on the whole dependency graph, including “implicit” inputs, allowing for transformations not
<pie_> they are doing all the stuff i keep wishing we had
<gchristensen> nice
<pie_> why why why
<pie_> dot gif
<gchristensen> once I tried to get cache.n.o to have by-default separated debug info, but it would be like 10x increase in disk usage per evaluation
<lovesegfault> pie_: Yeah, it's _really_ cool
<JJJollyjim> oof gchristensen :(
<JJJollyjim> debian pretty much does it at this point, i love it
<JJJollyjim> ubuntu too i think
<samueldr> they have far less to contend with with regards to output size, when a library closer to the root of stdenv is rebuilt, they generally can just update *that* library and not the whole world
<pie_> install guix on nixos
<JJJollyjim> yeah:(
<JJJollyjim> i'm worried about how quickly my laptop's SSD is gonna die when i have to overwrite basically my whole closure on every update
<samueldr> I actually wouldn't (and don't) concern myself with that
<samueldr> SSDs are more resilient than they were when this was actually a thing to look-out for
<JJJollyjim> aren't the MLC SSDs with more and more layers less resiliant?
<samueldr> than the non-layered ones yes
<samueldr> but more resilient than old-school SSDs when it was an issue
<samueldr> [citation quite needed]
<V> I really need to move my laptop to ZFS
<V> unfortunately, no backup machine, nothing to use if I screw it up, and I need to get extra disks to move all the data to while I make the switch
<pie_> thats a constant problem i have
<pie_> i just simply cannot manage to unfork my data
<V> actually that's a lie, I do have another machine
<V> just don't fork it, then :p
tilpner_ has joined #nixos-chat
tilpner has quit [Remote host closed the connection]
tilpner_ is now known as tilpner
endformationage has quit [Quit: WeeChat 2.9]
<energizer> do computers still beep when they turn on? (i'm having boot issues)
pie_ has quit [Ping timeout: 272 seconds]
pie_ has joined #nixos-chat
<lovesegfault> energizer: laptops don't
<lovesegfault> some desktops do
waleee-cl has quit [Quit: Connection closed for inactivity]
<ashkitten> if you attach a pc speaker
<cole-h> elvishjerricco: And the last cksum error disappeared... 👀
<elvishjerricco> That's a Halloween story right there
<elvishjerricco> cole-h: ^
<cole-h> lol 100%
pie_ has quit [Ping timeout: 246 seconds]
christophh is now known as Christophh
pie_ has joined #nixos-chat
<sphalerite> samueldr: they also don't keep all historical versions of everything I think.
lopsided98 has quit [Ping timeout: 264 seconds]
lopsided98 has joined #nixos-chat
cole-h has quit [Ping timeout: 260 seconds]
Christophh has left #nixos-chat ["User left"]
supersandro2000 has quit [Quit: The Lounge - https://thelounge.chat]
supersandro2000 has joined #nixos-chat
thibm has joined #nixos-chat
<supersandro2000> note: currently hard linking saves 9010.08 MiB
<supersandro2000> 140234 store paths deleted, 178437.74 MiB freed
<supersandro2000> might have build one or the other PR
<V> wew
<sphalerite> infinisil: does {^_^} keep track of GC records yet? :p
__monty__ has joined #nixos-chat
supersandro2000 has quit [Ping timeout: 264 seconds]
supersandro2000 has joined #nixos-chat
pie_ has quit [Ping timeout: 240 seconds]
<infinisil> Hehe not yet
pie_ has joined #nixos-chat
<gchristensen> 2567391 store paths deleted, 62960.56 MiB freed
<gchristensen> 9599 store paths deleted, 680032.26 MiB freed
<gchristensen> first one is hydra.n.o, Taneb got the second one
<sphalerite> gchristensen: is there a digit missing from the MiB in the hydra one?
<sphalerite> or is hydra just the record for path count and Taneb's for size?
<gchristensen> path count and size
<gchristensen> for hydra it had GC'd itself regularly to keep space, but was running out of inodes so it had to delete a zillion drvs
<sphalerite> ah right
<sphalerite> I'm somewhat disappointed we haven't had a terabyte GC reported here yet :p
<sphalerite> We should gc cache.nixos.org :D
<sphalerite> (disregarding the feasibility issues)
<gchristensen> *clicks the delete button*
lassulus has quit [Ping timeout: 256 seconds]
lassulus has joined #nixos-chat
<andi-> good way to check if we can still bootstrap without amazon
<gchristensen> lol I mean it would be a way
<andi-> I guess the answer would be No as our bootstrap binaries are on a bucket there ;)
<Taneb> sphalerite: at my previous job we had some... very large nix builds
<Taneb> Into the 10s of GB, which I expect are fairly unusual on h.n.o
<Taneb> This should account for the discrepancy in the ratios
<gchristensen> ow
<sphalerite> fair enough
<gchristensen> have you lost the plot if you're generating .json files as input to terraform, and barely use the actual terraform syntax?
<adisbladis> No, the terraform syntax is terrible
<adisbladis> Generating json is the sane option
<gchristensen> :D
<thibm> =
thibm has quit [Ping timeout: 256 seconds]
<andi-> gchristensen: reloading sway just crashed it -.-
<andi-> Still better than the gnome/kde experience where your session is restarted on every update..
<gchristensen> tragic and rare :(
<supersandro2000> andi-: ob Debian I just update half the plasma and get a little notification that I should restart.
thibm has joined #nixos-chat
<supersandro2000> Windows vibes
<andi-> On Debian you do ugly things to packages.
<andi-> and ugly things to get packages.
<sphalerite> andi-: yesterday I saw in some production code `wget http://archive.ubuntu.com/…/grub….deb && dpkg -i grub….deb` :|
<andi-> works.
<andi-> (most of the time)
<sphalerite> though really it's ok because they did apt-get -f install afterwards
<infinisil> > MiB
<{^_^}> 1048576
<infinisil> 680032.26 * MiB
<infinisil> > 680032.26 * MiB
<{^_^}> 7.13066e+11
<etu> > MB
<{^_^}> 8000000
<etu> :D
<etu> Good that both exist
<sphalerite> > megabyt
<{^_^}> undefined variable 'megabyt' at (string):434:1
<sphalerite> > megabyte
<{^_^}> undefined variable 'megabyte' at (string):434:1
<sphalerite> > mega * byte
<{^_^}> undefined variable 'mega' at (string):434:1
<sphalerite> > M * byte
<{^_^}> undefined variable 'M' at (string):434:1
<sphalerite> noooooooo
<sphalerite> infinisil: it's not an adequate pkgs.units replacement yet ;)
<infinisil> Hehe
<infinisil> Yeah there's just the abbreviated versions for now
<infinisil> Mb, MB and MiB (and the same for k*, G*, ...)
supersandro2000 has quit [Ping timeout: 246 seconds]
<sphalerite> > b
<{^_^}> "asdfghjkl"
<sphalerite> > Mb
<{^_^}> 1000000
<sphalerite> lol
<sphalerite> > :v Mb
<{^_^}> Mb = 1000 * kb
<sphalerite> > :v kb
<{^_^}> kb = 1000
<sphalerite> wait, they're not units :o
<sphalerite> > :v m
<{^_^}> m = dimUnit.l
<sphalerite> > dimUnit
<{^_^}> { i = <CODE>; j = <CODE>; l = <CODE>; m = <CODE>; n = <CODE>; o = <CODE>; t = <CODE>; }
supersandro2000 has joined #nixos-chat
<infinisil> sphalerite: bytes are unit-less :)
<infinisil> It's just a number really
<infinisil> (the amount of bits)
<infinisil> > :v kB
<{^_^}> kB = 1000 * byte
<infinisil> > b = 1
<{^_^}> b defined
<infinisil> > B = 8
<{^_^}> B defined
<infinisil> > kB = 1000 * B
<{^_^}> kB defined
<infinisil> > ZiB = 1024 * EiB
<{^_^}> ZiB defined
<infinisil> > 17592186044416 * MiB / ZiB
<{^_^}> division by zero, at (string):51:100
<infinisil> > B = 8.0
<{^_^}> B defined
<infinisil> > 17592186044416 * MiB / ZiB
<{^_^}> division by zero, at (string):51:100
<infinisil> > 17592186044416.0 * MiB / ZiB
<{^_^}> division by zero, at (string):51:100
<infinisil> Damnit!
<infinisil> I'll go to #bottest
<infinisil> > 17592186044416.0 * MiB / ZiB
<{^_^}> 0.015625
<infinisil> There!
<infinisil> sphalerite: ^
<infinisil> Redefined KiB to be a float, which then propagates all the way up :)
<sphalerite> infinisil: but bits are a unit!
<infinisil> Are they?
<sphalerite> Sure
<sphalerite> Does it make sense to multiply 8 bits by 8 bits?
<sphalerite> Would the result be in bits too?
<sphalerite> I'd consider the result to be in bits^2
<joepie91> it doesn't make sense to multiply bytes by bytes either
<joepie91> doesn't make it not a unit :P
<infinisil> Hmm I'm not sure
<infinisil> Do we need units for everything that can be counted?
<joepie91> far as I know multidimensionality or w/e you call it isn't required for something to be a unit
<thibm> It's not only about counting, it's also about composing units. Think about b/s or B/s
<sphalerite> thibm: well you could consider that as just s^(-1) or Hz ;)
<thibm> I don't really agree
<sphalerite> I do prefer considering information as a dimension myself, just trying to think of how one could argue against it :)
<thibm> In my head, b/s is not of the same dimension than a frequency (even in a broad sense), but I admit that I cannot really explain why :p
<thibm> bits are really important in information theory
<gchristensen> tumblr's in-house metrics dashboard before they ditched it for prometheus measured stuff in Hz and it was the absolute worst thing
<sphalerite> as in "global outages occur at a rate of 23nHz?"
<gchristensen> lol yeah
<sphalerite> hahaha
<thibm> One could argue that you need to add 'binary-units' option to latex's siunitx package to get \bit and \byte ;) (you also get \mebi and others. I don't get why it's not provided by default)
<gchristensen> maybe my PR will get a review on day 2 https://github.com/hashicorp/terraform-provider-aws/pull/16373
<{^_^}> hashicorp/terraform-provider-aws#16373 (by grahamc, 22 hours ago, open): aws_ebs_snapshot_import: init resource (the red X is from a panic() in a test which I don't know how to avoid)
<andi-> Do we have a nixos related markdown pat service running somewhere that will not disappear soon(tm)? :D
thibm has quit [Ping timeout: 256 seconds]
supersandro2000 has quit [Ping timeout: 264 seconds]
thibm has joined #nixos-chat
FRidh has joined #nixos-chat
pie_ has quit [Ping timeout: 240 seconds]
pie_ has joined #nixos-chat
<gchristensen> colemickens: btw you might be in to #nixos-cloud as a sort of cloud working gorup
waleee-cl has joined #nixos-chat
supersandro2000 has joined #nixos-chat
red[evilred] has joined #nixos-chat
<red[evilred]> 23nHz is the best unit I've ever heard of
<Taneb> I think it's roughly an event every year and a half
<immae> it looks boringly flat
<eyJhb> I didn't even know that they had Avengers LEGO until I got some. Maybe that makes me a bad employee
<Taneb> Hmm, I wonder if nBq is a better unit for outages
<aanderse> avengers lego... uh oh :S
<Taneb> It's unlikely they're periodic
<eyJhb> aanderse: I am not the biggest fan, but I do like LEGO however :p
<aanderse> probably not best to start talking about lego around me 😧 i might not shut up
<aanderse> ha ha ha
<sphalerite> $ units '1/(23 nHz)' 'year'
<sphalerite> * 1.3777724
<sphalerite> / 0.7258093
<sphalerite> red[evilred]: Taneb ^
<red[evilred]> so aanderse (IRC) is an afol eh? ;-)
<sphalerite> once every 1.38 years, or 0.726 times per year
<eyJhb> aanderse: You a big LEGO fan?
<aanderse> oh boy... i better not start posting pics :P
<eyJhb> Do it, do it, do it!
<red[evilred]> I want to see your organized lego skills
<eyJhb> I only have one piece, that I think is "rare" for some
<red[evilred]> followed by your declarative lego model specification ;-)
<eyJhb> But that is also quite lovely
<sphalerite> can we generate ldraw from nix?
<aanderse> oh boy, i'm going to have to go take a pic...
<eyJhb> "a pic", that better be a big one aanderse !
<eyJhb> Ohh cool sphalerite , didn't know we had that!
<eyJhb> I am being worse and worse by the minut, Also, I have no dedicated shelf
<sphalerite> eyJhb: units or ldraw?
<eyJhb> ldraw :p
<eyJhb> Did we get all the units in the bot?
<aanderse> so not strictly avengers, but spiderman is an avenger every once in a while, so related: https://images-na.ssl-images-amazon.com/images/I/71yep9yufWL.AC_SX425.jpg
<aanderse> the spiderman bridge set
<aanderse> red[evilred]: eyJhb either of you familiar with that one?
<eyJhb> No :(
<eyJhb> But I know the bridge, didn't know they had a set :p
<aanderse> yeah, its a neat little set
<aanderse> about 1000 pieces
<aanderse> but it doesn't have that "wow" factor
<aanderse> unless you put 4 of them together: https://imagebin.ca/v/5iSwNf84dqJ4
<aanderse> >_<
<aanderse> you really need to see it to get the full effect, and its covered in figures right now, but yeah in its full glory it is a sight to see :D
<aanderse> the avenger hellicarrier sits right beside that bridge currently
<joepie91> what a surprise, this PR hasn't been merged yet: https://github.com/jonschlinkert/get-value/pull/21
<{^_^}> jonschlinkert/get-value#21 (by joepie91, 13 weeks ago, open): Correct dot-key claims
<joepie91> I sure wonder why /s
<joepie91> (this is mr. benchmark cheater)
<__monty__> How many spidermen does it take to defend a bridge? : )
evanjs has quit [Quit: ZNC 1.8.2 - https://znc.in]
FRidh has quit [Ping timeout: 240 seconds]
evanjs has joined #nixos-chat
FRidh has joined #nixos-chat
thibm has quit [Quit: WeeChat 2.6]
cole-h has joined #nixos-chat
<aanderse> well there is spiderman, scarlet spider, spiderwoman... and then the other spiderman versions i added
<aanderse> so quite a few, in their opinion
<aanderse> well, my kids added
<sphalerite> joepie91: "my software is the best. See this data here." "You expect me to have time for updated data? I'm too busy being awesome!"
<joepie91> sphalerite: yeeeeeup.
<joepie91> I imagine he can't close this one for 'trolling' since I'm on my best behaviour, and so he just ignores it instead...
FRidh has quit [Ping timeout: 240 seconds]
FRidh has joined #nixos-chat
<gchristensen> anyone write really complicated jq expressions, struggle to find the operators you want, and ditch and rewrite it in nix?
<srk> I was toying with jq like query language for nix a bit
rajivr has quit [Quit: Connection closed for inactivity]
<hexa-> there is also gron https://github.com/tomnomnom/gron
<joepie91> gchristensen: JS for me usually, but yes
<V> nah I usually find jq is sufficient for what I do
FRidh has quit [Ping timeout: 240 seconds]
<eyJhb> aanderse: Cool!
<eyJhb> I had to do some shopping real quick, sorry for not anwsering. But damn that is nice! Taking the set and actually making it better :D
endformationage has joined #nixos-chat
<sphalerite> gchristensen: not really, either I make it work in jq or I get distracted and do something completely different :p
<sphalerite> tbf I haven't really written any super complicated jq queries, I think.
<sphalerite> jq is the most amazing complement for bash's structured-data weakness though.
Baughn has quit [Ping timeout: 244 seconds]
Baughn has joined #nixos-chat
<colemickens> I don't like when LICENSE files don't plainly identify themselves? like not everyone has MIT memorized?
<joepie91> yeah...
<samueldr> same :/
<joepie91> I usually work with JS and look in package.json instead :P
<joepie91> SPDX++
<samueldr> I select a huge chunk and search for it
<sphalerite> I get the MD5 hash of it and put it in google
<samueldr> sphalerite: it breaks for most MIT and BSD licenses where you add your own copyright information
<sphalerite> abathur: … no
<sphalerite> x)
<abathur> hehe
<sphalerite> abathur: OH but that isn't valid JSON!
<abathur> fair, I'm cheating
<abathur> but, I just updated it with a screenshot of jq formatting its output just fine
<sphalerite> x)
supersandro2000 has quit [Ping timeout: 265 seconds]
supersandro2000 has joined #nixos-chat
<sphalerite> $ bluetooth
<sphalerite> bluetooth = on
<sphalerite> lol
__monty__ has quit [Quit: leaving]
red[evilred] has quit [Quit: Idle timeout reached: 10800s]
<andi-> anyone else seeing this issue? https://github.com/grafana/loki/issues/2660
<{^_^}> grafana/loki#2660 (by mweinelt, 8 weeks ago, open): promtail uses previous unit label for dmesg messages
<JJJollyjim> hmm don't think so
<JJJollyjim> oh i don't use unit, only syslog identifier
<JJJollyjim> it seemed nicer for some reason, can't remember why
<JJJollyjim> i'd like to collect more labels from journald but don't want to increase the cardinality
<JJJollyjim> maybe i should put them in json or something