<colemickens>
I am beyond myself with excitement over wfvm
<infinisil>
Damn that's neat
<infinisil>
Not that I care about Windows, but still
<infinisil>
Maybe allows some other people to care less about Windows
<Ashy>
"A Nix library to create and manage virtual machines running Windows, a medieval operating system found on most computers in 2020. The F stands for “Functional” or a four-letter word of your choice."
<Ashy>
haha
<Ashy>
that's great
<drakonis>
wfvm
<drakonis>
my god
<abathur>
nice
rajivr has joined #nixos-chat
cjpbirkbeck has quit [Quit: Goodbye, take care]
<abathur>
speaking of the tag cloud for wfvm; is anyone up on an easy way to cross-platform-sandbox execution of arbitrary shell scripts? best I've got is downloading a docker image atm, and that's got an external dependency at least on macos
<aleph->
Lot of neat stuff this week
<drakonis>
highly
Taneb has quit [Quit: I seem to have stopped.]
Taneb has joined #nixos-chat
waleee-cl has quit [Quit: Connection closed for inactivity]
<abathur>
heh
<samueldr>
somtimes I don't understand git bisect's output
<samueldr>
Bisecting: 3009 revisions left to test after this
<samueldr>
for the two last bisect actions
<ashkitten>
oof, my server's been scrubbing its zpool for almost 18 hours now and is 55% done
<ashkitten>
i *need* a better server
cole-h has joined #nixos-chat
endformationage has quit [Quit: WeeChat 2.7.1]
<samueldr>
that's a long time
<samueldr>
I just shutdown mine, open the HDDs and scrub each platter with a toothbrush
<samueldr>
(people: don't)
<samueldr>
(it's not a good idea to brush your platters)
<cole-h>
What about *air*brushing?
<ashkitten>
it's... generally not a good idea to unseal your hdd :p
<ashkitten>
my desktop ones are filled with helium
<ashkitten>
probably would perform worse with that replaced by air
<ashkitten>
(also, the drives are sealed in a dust-free environment. your home is not dust-free)
<Arahael>
I imagine HDD's are only far more sensitive to dust today than they were two decades ago. And it wasn't a good idea to open them even back then.
<sphalerite>
ashkitten: how big is the pool?
<ashkitten>
1TB.. i forgot i accidentally still have auto snapshots on
<ashkitten>
oops
<sphalerite>
oh ok, yeah, >18h for 1TB really is slow…
<ashkitten>
yeah
<ashkitten>
it's uh
<ashkitten>
really bad
<ashkitten>
i don't have a slog
<ashkitten>
because it's a kimsufi server i have no control over hardware-wise
<sphalerite>
slog won't affect scrubs though?
<elvishjerricco>
ashkitten: My 4TB USB drive scrubs in 6 hours
<elvishjerricco>
50% full
<elvishjerricco>
What ZFS version are you using?
<ashkitten>
sphalerite: it'll affect scrubs if there's constant writes..
<elvishjerricco>
0.8 heavily optimized scrubbing by scanning the disk so it can do the bulk of the reads sequentially
<ashkitten>
i know
<ashkitten>
my disks are slow and i have too much io
<sphalerite>
ashkitten: well the stuff in the slog will still need to be flushed to the disks
<ashkitten>
i'm running a mastodon server and a matrix server
<ashkitten>
the minimum latency i get in normal operation is 30ms
<sphalerite>
the only case where a slog _could_ (I'm not sure) reduce the amount of writes happening is if the same thing is being overwritten repeatedly
<ashkitten>
it's really bad
<sphalerite>
30ms doesn't seem that bad to me?
<ashkitten>
i will be moving to a new server with ssds soon
<ashkitten>
sphalerite: that's *minimum*
<ashkitten>
it's usually well over 100ms, i think
<sphalerite>
right, but minimum latency is most likely composed mostly of network latency
<ashkitten>
no, i mean zfs reported latency
<sphalerite>
oooooooh.
<elvishjerricco>
ashkitten: How do you get ZFS to report latency? That sounds useful
<ashkitten>
for reference, my desktop hovers around 2ms latency
<ashkitten>
elvishjerricco: prometheus and grafana
<elvishjerricco>
Ah
<ashkitten>
but you can also query it directly with procfs
jared-w has quit [Ping timeout: 244 seconds]
jared-w has joined #nixos-chat
<makefu>
for anyone who is still looking for some sweet sweet nixos stickers, there is currently a promotion by stickermule where you can get 10 die-cut stickers for 90 cents ($1), free shipping: https://www.stickermule.com/de/deals/87017a28?ref_id=2277921701 (ref link). die-cut means that they will cut out that sweet sweet lambda snowflake for you
cole-h has quit [Quit: Goodbye]
<ashkitten>
ohh wow
<makefu>
the quality is absolutely fantastic
<ashkitten>
how long is this for?
<FireFly>
:o
<eyJhb>
Uhhhh
<eyJhb>
Want
<ashkitten>
i might have to order like... a ton in various pride colors
<FireFly>
oh hmm, that's a good idea
<eyJhb>
But need to create a account :(
<FireFly>
there might be a max cap on the offer per person/order
<makefu>
i think there is a cap at 1 per account
<ashkitten>
aw
<eyJhb>
pr. account is key
<makefu>
(however policing one account per person is hard)
<eyJhb>
But not fair to abuse them
<eyJhb>
Or be like makefu :D
<ashkitten>
if i could get multiple color schemes per run thatd be cool
<makefu>
i won't do this. there is even a proofing-phase where a `real person` prepares your cut
<eyJhb>
Wonder if it is pr. person, or pr. household etc.?
<makefu>
not per household, that is for sure
<eyJhb>
I am sure my cat wants some as well then
<FireFly>
ashkitten: if you're okay with slightly smaller, you could presumably do a design with e.g. 4 different colourschemes scaled down, and get 10 copies of all 4
<ashkitten>
hm
<FireFly>
or otherwise try to make full use of the 74x74 mm :p
<FireFly>
76x76 *
<philipp[m]>
Who just managed to get a first server at work migrated to nixos for evaluation purposes? This guy!
<makefu>
i really like this. tabs i want to keep i just bookmark
<srk>
with chrome it's even worse as with too many tabs the tab bar collapses into favicons and then just close buttons after a while
<srk>
there's a tab search extension tho.. :D
<makefu>
"the only thing with so many tabs is to close them" is what the chrome devs thought
<eyJhb>
I actually enjoy that
<eyJhb>
Since most af my tabs are on a linear timeline, so everything before X can be closed 99% of the time :p
<eyJhb>
Hate the way FF does it
<Mic92>
Just close them automatically
<Mic92>
The rest is history
<eyJhb>
Never had tab open for a week :p
<bqv>
I really hate living in a city
<srk>
starting to have similar feelings
<FireFly>
hm, how so?
<bqv>
For me, its the constant noise pollution
<srk>
too hot, too noisy
<bqv>
The only way to ever get peace and quiet is to be up at 2am
<srk>
or headphones and music
<srk>
or loud music
<FireFly>
Oh
<bqv>
Other than that, your lawnmowers, leafblowers, chainsaws, screaming children
<bqv>
Noisy teenagers, motorbikes so loud they make your cutlery wobble
<srk>
this.
<bqv>
Trucks on gravel taking 20 minutes to park
<bqv>
I'm going insane I tell you
<etu>
Beep beep beep
<FireFly>
right, I wouldn't want to live centrally in a big city
* srk
lives in the quieter part of the city but lots of these as well
<etu>
I've spent time on the countryside during the summer and can confirm that lawnmowers exists there as well
<pie_>
bqv: earplugs are pretty great
<joepie91>
bqv: don't forget rumbling cars
<pie_>
i would have gone crazy in my high school dorm without them
<joepie91>
and car doors
<joepie91>
car doors are the bane of my fucking existence
<joepie91>
SLAM
<joepie91>
SLAM
<joepie91>
SLAM
<bqv>
Yes
<bqv>
Very much
<FireFly>
mm, student apartment was too noisy for me
<bqv>
pie_: if I manage to block out enough sound with a combination of earplugs, headphones, and white noise, I just start to hallucinate sounds instead
<sphalerite>
oh yeah student apartment… I lived in one for the first year of uni, and it was right next to a building site
<joepie91>
also just cars-- generally
<joepie91>
bqv: anyway I moved out of the city for this reason
<sphalerite>
where for some reason they were working nights and weekends too
<bqv>
Yeah
<joepie91>
bqv: some things remain (car doors, fucking leafblowers, etc.) but at least it's better, and I can go outside and go somewhere without feeling like I have to wade through a mass of people and noise first
<FireFly>
suburbs can be good, can get you both "walk to quiet forests" and "train to city"
<FireFly>
at least in my experience
<joepie91>
leafblowers should be banned
<joepie91>
FireFly: yeah that's sort where I live now
<joepie91>
town at the edge of a medium-sized city
<joepie91>
sorta*
<FireFly>
*nod*
<joepie91>
medium-sized by NL standards anyway :P
<bqv>
Yeah, I've had good suburb experiences tbh. I'm living on the edge of a tiny city now so I guess this doesn't qualify as a suburb
<bqv>
We have a car, I care less and less about being "well connected" by the day
<FireFly>
I don't drive though, and don't like cars much anyway :p
* joepie91
does not have a car
<joepie91>
but this is NL :P
<FireFly>
hehe
<joepie91>
I'm probably one of the few people here actually with a driveway without a car
<FireFly>
I've not really visited NL properly, only barely set my foot there :p
<bqv>
I mean I think I'd need it if we ended up moving full countryside
<FireFly>
*nod*
<joepie91>
but yeah, by bicycle, it's 5 minutes to the town center (which has nearly everything one might need), 10 minutes to the train station
<joepie91>
literally one corner to the dentist, 10 minutes to the vet
<joepie91>
and like 30 minutes to the city center of Den Bosch?
<joepie91>
30-40
<joepie91>
a car would be useful for long distances due to bad night coverage of public transport here, but otherwise, I can do just fine without one
<FireFly>
I don't own a bicycle, and haven't cycled in years, which is probably uthinkable of in NL terms :p
<FireFly>
+n
<FireFly>
(or almost so)
<FireFly>
I like walking though
<joepie91>
most people here have at least 2 bicycles
<joepie91>
yeah :P
<joepie91>
I think
<joepie91>
one at home, one at wherever their workplace is
<joepie91>
for the last mile from train to workplace
<bqv>
pie_: the ganzfeld effect, is what I was talking about
<joepie91>
or university
<joepie91>
etc.
<joepie91>
or a second mountainbike-y bicycle for leisure or w/e
<joepie91>
FireFly: seriously though, if you are ever in NL, get a bicycle. even if just here as a tourist
<eyJhb>
ANd don't spill any water, it might overfill NL with it :p
<joepie91>
they're generally pretty cheap to rent, but honestly, if you're staying in NL for like 2+ weeks it may be cheaper to just buy a second-hand one locally and literally give it away afterwards :P
<eyJhb>
As far as I remember, most of NL is below water level, or semthing
<joepie91>
lol
<joepie91>
I think about half of it is below sea level yeah
<eyJhb>
:D
<eyJhb>
No spilling!
<bqv>
I did love how bikes were first-class road citizens when I went there
<eyJhb>
But yeah, NL is a lovely place
<bqv>
And at least theyre mostly silent…
* bqv
moves to ban cars! But you go first.
<eyJhb>
When I went there, it was on a exchange trip
<FireFly>
joepie91: maybe :p
<joepie91>
https://www.youtube.com/c/BicycleDutch/videos has some interesting videos about bicycle infrastructure here, actually taught me a lot that I didn't know yet, and I live here!
<eyJhb>
At that time, it was popular with having your "pants" in your socks. Which apparantly meant "I WANT TO FIGHT" at the exchange school we were at..
<FireFly>
assuming MCH happens >.>
<joepie91>
FireFly: hopefully!
* FireFly
. o O ( putting the 'may' in MCH )
<joepie91>
but yeah, cannot recommend above videos enough
<FireFly>
I should watch 'em later :3
<joepie91>
(first channel does have a lot of just bike ride videos, I don't know if there's a separate playlist for the infrastructure stuff)
<joepie91>
FireFly: etu: speaking of which, as someone involved in the volunteer orga for MCH: have you volunteered at one of the Dutch hacker camps before, and if no, what would motivate you to do so?
<joepie91>
(same question for anyone else who has been / wants to go :P)
<FireFly>
no, but I also haven't attended any of the dutch camps before :p
<FireFly>
so I might, we'll see, depends on my mood at the event probably..
<etu>
I've only been to one dutch camp... So don't have any good answers. Been to more of the CCC events (including 2 camps)
<joepie91>
etu: nono, that means you're precisely the right audience for this question :) the 'regulars' are easy, those are generally involved with volunteering already, the goal is to make it more accessible and attractive for people who *don't* already have close ties with the 'scene'
<etu>
joepie91: It takes a fair bit of organization just to plan the roadtrip from Stockholm and back :D
<joepie91>
FireFly: is there anything the volunteer orga could do to make it more likely / easier for you to volunteer? like avoiding certain kinds of stress or responsibilities, creating a certain environment, that sort of thing
<f0x>
I got my close ties with the 'scene' by volunteering at SHA :P
<bqv>
srk: for twitch I use mpv+irc :p
<etu>
joepie91: I guess the "standard" perks applies, like "some food" and access to different color of merch or something like that?
<bqv>
Works for everything except discovery…
<joepie91>
f0x: lots of people don't though, so it's worth looking outwards to better understand why people would / would not be motivated to get involved by themselves
<f0x>
oh absolutely
<srk>
bqv: yeah, mpv too, haven't configured irc yet but I will :) less heat wasted
<joepie91>
etu: both of those are pretty much already a certainty, at least. high-quality volunteer food and volunteer t-shirts (if you do a certain minimum amount of volunteering) are AFAIK already on the list of must-haves - are there any other things you could think of?
<joepie91>
(we should probably make sure to clearly communicate the food etc.)
<etu>
joepie91: For me I have a bit of a difficult relationship with my employer regarding conferences. They don't believe in education. So I have to spend vacation. So it becomes a bit mixed for me... "vacation + work is weird"
<etu>
joepie91: But I also know that many people say it's awesome to volunteer at such events
<joepie91>
etu: do you think that could be addressed by making walk-in volunteering a thing? in that you can also just walk up to the volunteer tent or some random team whenever you feel like it and ask "hey do you have anything to do for me right now", and it would be counted as the usual volunteer work? just without a pre-planned 'shift' and without a guarantee that a shift will be available for you
<etu>
joepie91: Because of the friends you can make, and extra things you can learn from people you wouldn't meet otherwise. So I haven't actually done much volunteering because of this conflict in my mind... I've thought of volunteering at congress a couple of times, but then it's been too late and no shifts are available, etc.
<FireFly>
I'm hoping to volunteer, I did a tiny bit at emfcamp
<etu>
joepie91: Yes, that could probably work for me.
<FireFly>
also yeah, that'd probably work well for me too
<eyJhb>
Mostly I have done it, to just be a part of it, being behind the scenes etc. is always fun. Food as stated is a must, and having clothes that makes you part of the crew community as well
<FireFly>
the stuff I did at emf was kinda "walk in and see what's needed"
<joepie91>
FireFly: etu: so another thing is the 'flying volunteer', where basically you have a 'shift' at the volunteer tent where you can just hang around and do whatever until some misc task shows up on the board for someone to do and/or fill in for someone with a full shift who hasn't shown up, as a sort of 'spare pool' of all-round volunteers... what would your thoughts be on that?
<joepie91>
(yes this has turned into one of those survey forms :P)
<eyJhb>
joepie91: we have had that, but that was mostly just people who hang around anyways
<joepie91>
eyJhb: right, here is the intention to really make if a shift
<joepie91>
eyJhb: another idea that's under consideration is having the same thing, except instead of a shift at the volunteer tent you can just freely walk around and do your thing and get a push notification or w/e when there is something to do
<joepie91>
(except that wouldn't be counted as volunteer hours unless you actually pick up one of the jobs, ofc)
bqv has quit [Quit: WeeChat 2.9]
bqv has joined #nixos-chat
<bqv>
bizarrely, enabling bcachefs breaks btrfs, for me
__monty__ has joined #nixos-chat
<eyJhb>
joepie91: Is there any reason it shoud not be conted? What is the bonus of ++ vulunteer hours?
<eyJhb>
bqv: when will you get ZFS?
<talyz>
bqv: there's only room for one cow in the kernel
<joepie91>
eyJhb: food, volunteer shirts, etc.
<joepie91>
basically you get volunteer stuff if you volunteer
<eyJhb>
Does hours convert to food etc.?
<eyJhb>
Yeah, but does 1 hour equal the same as 1000?
<joepie91>
and having 'roaming volunteers' who don't get counted unless they actually pick up a job means the pool of 'spare volunteers' can be expanded a lot without either significantly increasing volunteer costs nor 'cheapening' the dedicated hours that non-roaming volunteers put in
<joepie91>
eyJhb: sort of. it's more like "if you have done meaningful volunteering, then you get access to X", it's not like 1 hour = 1 plate of food
<eyJhb>
Damn... 8 hours work = lots of food
<eyJhb>
But it is difficult to do
<cransom>
tapas. very small plates.
<eyJhb>
The stand by is maybe overkill
<joepie91>
the thing is that there's always going to be a lot of volunteer work that requires staying on-site and not being able to go and do whatever the hell you want, which is part of the reason we want to reward people who put in that work
<eyJhb>
cransom: NO!
<joepie91>
but that would be rather cheapened by giving anyone the reward just for ticking a box
<joepie91>
even leaving aside the economics
<eyJhb>
joepie91: How many volunteers do you need, cirka?
<eyJhb>
circa?
<eyJhb>
Whatever, ca.
<joepie91>
hence why roaming volunteers would not be counted unless they do work, as it involves no commitment :)
<joepie91>
eyJhb: all of them
<joepie91>
quite literally
<joepie91>
the Dutch hacker camps are very much built on an "everyone pitches in" idea
neeasade is now known as edasaeen
<eyJhb>
You need to divide it up then
<bqv>
eyJhb: i will never use zfs
<bqv>
talyz: lol
edasaeen is now known as neeasade
<joepie91>
also there will actually be maximum volunteer hours :)
<joepie91>
too many people who will burn themselves out otherwise
<eyJhb>
There is a basis of, everyone helps everyone, which you just do, then you have people who actually need to be in charge of X before they can help
<eyJhb>
What duration are we speaking?
<joepie91>
(orga included)
<eyJhb>
We did 320+ people for three days, 5 people
<eyJhb>
24/7 event :p
<joepie91>
eyJhb: exact details are up in the air, right now the list basically just says "we want to limit hours to something that doesn't burn people out"
<joepie91>
it's not going to be something stupid like an hour a day
<joepie91>
but we're also not going to let you do shifts from wake to sleep
<eyJhb>
But how many days is the event for?
<joepie91>
~4
<eyJhb>
Including setup, or is that ~4 guest days?
<joepie91>
that is 4 public days
<joepie91>
I think including setup is like 2-3 weeks total?
<joepie91>
not sure though, not my department :P
<eyJhb>
That seems like a loooong time
<joepie91>
<eyJhb> The stand by is maybe overkill
<joepie91>
probably won't be
<joepie91>
this is a massive event with a lot of work
<joepie91>
we're talking 3000+ visitors here
<joepie91>
on a huge terrain that you pretty much need a golf cart to get around on, when doing jobs
<eyJhb>
But, there is a need to split the tasks up, what needs to be done and who can do it. This sits some constraints on who can do what, and what "rank" one might be, I would say... Because, needing everyone is not a good thing, you can get too many, and you don't want them to feel insignifinat I guess
<eyJhb>
But having standby is OP
<joepie91>
eyJhb: all this has been considered :)
<joepie91>
with "everyone" I mean that we *want* everyone to be a part of it, rather than it being a handful of people running it centrally
<joepie91>
rather more people doing less hours than the inverse
<eyJhb>
Maybe some enjoy doing many hours? :p
<joepie91>
(this has been a part of these camps for a long time)
<joepie91>
eyJhb: yes, that is part of the problem
<joepie91>
because there's a bunch of people who burn themselves out every camp
<joepie91>
it's not okay to need two weeks to recover from your 'holiday'
<eyJhb>
Well...
<eyJhb>
Do they really want to do many hours then, or do they feel forced?
<eyJhb>
E.g. I would mostly join such a event, just to be behind the scenes doing stuff
<eyJhb>
Joined a LAN party group, while not playing games or eating junk sums it up :p Just doing the servers
<joepie91>
<eyJhb> Do they really want to do many hours then, or do they feel forced?
<joepie91>
the line between these two is often difficult to see
<joepie91>
but in all honesty this isn't really something that's up for discussion, so I don't want to spend too much time discussing it :P
<eyJhb>
Either way you do it, you might scare some away :p
<eyJhb>
How many are actually arranging it?
<eyJhb>
And not really discussion, just saying what I know from experience
<drakonis>
there's a headless service module at nixpkgs
<drakonis>
has existed for a while
<bqv>
i've moved it from weechat-headless to a dtach'd weechat on tty8
<bqv>
yeah, i avoided that cause it's not very customizable
<bqv>
i think it hardcodes screen or something
<bqv>
but if that works for you /shrug
<bqv>
cransom: i was having a chat with arcnmx, it's actually pretty usable and mergeable as it stands
<bqv>
i considered using it, can't remember why i didn't
<drakonis>
ah yes, screen.
<drakonis>
it still does that, yes.
<eyJhb>
drakonis: Come join Matrix instead? :o
<eyJhb>
I so want to set it up, and then use their clients.
<bqv>
yeah, so if that doesn't work for you, or you want to use weechat-headless, you might as well just make your own service
<eyJhb>
Might be a lot smarter/better than Weechat relay
<drakonis>
nah man
<bqv>
eyJhb: almost certainly not
<drakonis>
matrix has issues with bridges everywhere
<eyJhb>
What issues?
<bqv>
lol
<eyJhb>
There are some, but most I have seen comes from using the official bridges
<eyJhb>
Haven't really noticed much from joepie91 , worldofpeace , etc.
<bqv>
so otherwise you have to set up your own entire matrix homeserver
<bqv>
that's way more painful and way more of an investment than weechat
<bqv>
i mean i love matrix, but as an irc bouncer, no.
<eyJhb>
Once they have a Messenger brigde, I am all in
<eyJhb>
But also, just generally it seems nice
<bqv>
facebook?
<eyJhb>
But not sure if weechat relay works with multiple clients (well)
<eyJhb>
Yeah
<eyJhb>
I use Bitlbee atm.
<eyJhb>
But I would like it to work "better"
<eyJhb>
ALso, the IRC clients for Mobile really suck
<bqv>
i'm more upset that you want to bring facebook into this ecosystem
<drakonis>
the bridges do drop a whole bunch
<eyJhb>
Do drop a whole bunch?
<bqv>
it's just overcomplicating things, imo. get weechat, job done
<eyJhb>
But, isn't weechat relay just complicated ZNC?
<eyJhb>
:p
<cransom>
i never /want/ to irc while mobile, but weechat + glowing bear works in a pinch if it's otherwise not possible.
<eyJhb>
cransom: Me neither must of the time
<bqv>
no, it's an irc client, that happens to allow you to turn on a bouncer with literally one line
<eyJhb>
The only reasons are when I have IRC + Messenger
<eyJhb>
Because I would rather use that, then the app
<bqv>
even znc overcomplicated things for less gain, the way i see it
<f0x>
eyJhb: my bridge tries to be a bit less annoying than the .org one :)
<eyJhb>
f0x: how come? And how so?
<f0x>
eyJhb: notable changes are a way shorter link to uploaded images/files, instead of a long action thing, and only using the auto-pastebin on 10+ lines instead of 3+
<f0x>
eyJhb: oh and no [m] default suffix
rajivr has quit [Quit: Connection closed for inactivity]
<eyJhb>
That is all config stuff, right?
<eyJhb>
I also need a short and sweed .dk domain name
<bqv>
also, there's weechat-matrix, so you can actually use weechat as a matrix client
<bqv>
i do that
<bqv>
so i can matrix in emacs
<bqv>
the only downside is the attachments, which you just reminded me of
<bqv>
if they're encrypted, they don't come through quite well
<bqv>
but that seems WAY more sensible to me than having irc in matrix
<bqv>
that way when matrix inevitably shit's the bread, irc still works
<f0x>
eyJhb: for the file upload things I had to dig in the code
<joepie91>
this is one of those things I wrote as a one-off, and now it turns out to be useful really often!
<joepie91>
for more-or-less atomic values
<joepie91>
/ operations
<__monty__>
Sounds a bit like linear typing : )
<samueldr>
__monty__: linear typing is when you make a red/black tree?
<samueldr>
who knew that red/black trees were cherry trees!
<__monty__>
Hmm, that joke went right over my head.
<samueldr>
cherry mx key switches, red and black are linear
<__monty__>
-_-
<samueldr>
you know a joke's good when it makes you cry, peeling layers off when explaining it, like an onion
<__monty__>
Red/Black trees give me PTSD.
<joepie91>
__monty__: I do apply linear typing principles in quite a few places
<__monty__>
Tried wrangling them into a type-safe API using dependent types.
Church- has joined #nixos-chat
cjpbirkbeck has joined #nixos-chat
<sphalerite>
adisbladis: was it you who recommended prisjakt.nu? Or some other Swede? :p
<sphalerite>
it came in handy just now in any case
<cole-h>
ZFS wizards: how do I reclaim storage from a deleted snapshot? I made a mistake when copying a rather large file and even after deleting both the file and snapshot, I have very little free space
<cransom>
if the data isn't referenced, it should be reclaimed automatically. you can speed it up with a `sync`
<joepie91>
__monty__: emphasis on "principles" - I'm not trying to exactly emulate existing type systems in JS, rather I am taking their principles and working out a sensible way to use them in JS without giving up on its properties of abstraction flexibility
<cole-h>
cransom: How do I make sure it's not referenced? `zfs list -t all -r rpool` doesn't show what is eating up all that space...
<drakonis>
nyxt...
<__monty__>
Static typing doesn't require giving up flexibility.
<bqv>
drakonis: it's cool, i just wish they'd make a release sometime this side of the century
<bqv>
i wonder if i could make github links redirect to magit-forge...
<cransom>
cole-h: if there aren't any snapshots or bookmarks, i'd expect it to reclaim it.
<joepie91>
discussion I don't feel like having tonight :)
<drakonis>
i need some advice regarding overlays
<drakonis>
really need that acme patchset now
<cole-h>
Grrr... gonna reboot.
cole-h has quit [Quit: Goodbye]
<bqv>
drakonis: if you need it desperately, you could use that branch as your nixpkgs
<bqv>
just means if it suffers from the standard PR lifetime in nixpkgs, you might end up getting pretty behind
<drakonis>
i'm aware of the pr lifetime
<drakonis>
nixpkgs churns 500 commits a week
<bqv>
i'm currently using an apparmor PR as my nixpkgs, actually
cole-h has joined #nixos-chat
<bqv>
cause i'm fairly optimistic about it
<eyJhb>
Or you can patch it using the most hacky way ever :D
<bqv>
the problem with both PRs mentioned is that they modify a lot of files
<bqv>
so not ideal
<bqv>
i do patch in two other PRs, currently, cause they're 1 file each
<__monty__>
joepie91: Same sentiment here. All I can do is ask you not to discount type theory so quickly.
<drakonis>
hmm, i think basalt should be able to handle flakes
<joepie91>
(I am not even discounting type theory at all)
<joepie91>
__monty__: I am not discounting it "quickly" :)
<joepie91>
(but static typing and type theory are not the same thing)
<cransom>
cole-h: based on the larg refer, i assume your big file is in rpool/local/nix?
<cole-h>
No, it was in rpool/system/var
<sphalerite>
cole-h: sometimes it takes a while to reclaim the space
<cole-h>
sphalerite: How long is "a while"? Because I'd really like to have that space back so I can re-copy that large file from a previous snapshot.....
<sphalerite>
cole-h: though I guess 742G is bigger than you want rpool/system/var to be?
<cransom>
cole-h: and you deleted any snapshot that ever held that file?
<__monty__>
joepie91: Clearly you're using a different definition of type theory. I'm consider in the study of type systems. Runtime tagging is cool but it's not a type system.
<joepie91>
<joepie91> discussion I don't feel like having tonight :)
<sphalerite>
cole-h: try `zpool get freeing rpool`, if that's 0 then yeah it's still references somewhere
<cole-h>
cransom: Only that ever held the mis-copied file... I was trying to restore a file from a previous snapshot by using `cp /var/.zfs/snapshot/snap/lib/libvirt/images/windows10 /var/lib/libvirt/images/windows10.raw`
<sphalerite>
cole-h: not really a solution to your problem, but have you considered using zvols instead of files?
<sphalerite>
s/references/referenced/
<cole-h>
sphalerite: Well, the situation is basically this: the pre-win10-resize holds a snapshot before I resized my win10 vm image (shrunk). All the subsequent snaphots are just after the VM was started basically. So I don't know if there's a way to remove all references, especially if I want to keep the VM image around...
<cole-h>
And yeah, `zpool get freeing rpool` is 0 :(
endformationage has joined #nixos-chat
<sphalerite>
ok, at this point I don't really get what you're trying to do or how you imagine it could be achieved
<cole-h>
sphalerite: Sorry, let me try to explain better.
<cransom>
if the snapshot is the .raw file, you can copy that from your last desired zfs snapshots and then destroy all of the snaps.
<sphalerite>
cransom: I guess the problem with that is that there's not enough space to copy it
<cole-h>
I only have 26.5G avail
<cole-h>
Whereas the image is in the hundreds of GB
<sphalerite>
is it compressible at all? :p
<sphalerite>
you could roll the filesystem back to the pre-resize snapshot, though that will roll back _all_ of /var.
<cole-h>
Why can't I roll it back to rpool/system/var@2020-08-09T220000? Would that not work?
<sphalerite>
sure it could work but it wouldn't help
<sphalerite>
I think?
<sphalerite>
You're trying to keep the pre-resize version around, right?
<cole-h>
That snapshot is prior to my mistake (which occurred this morning)
<cole-h>
Yeah, ideally. But I guess it doesn't matter anymore since I've dealt with any fallout from that..
<sphalerite>
then yes, that could be what you want. Use zfs diff to make sure you're not rolling back other changes in /var that you want to keep
<sphalerite>
(first thing that comes to mind that will probably be in there is the journal)
<sphalerite>
(but maybe that doesn't matter)
<sphalerite>
another idea: you could clone the pre-win10-resize snapshot, remove everything other than the disk image from the clone filesystem, then promote the clone… that could be funky though.
<sphalerite>
(then after promoting the clone destroy the old snapshots)
<sphalerite>
idk
<sphalerite>
but yeah move it to a zvol and snapshot it separately from the rest of /var, it will save you a lot of headaches :D
<sphalerite>
ok, I don't get why the space is still being eaten.
<cole-h>
Is it because I didn't use `-d`?
<sphalerite>
for what?
<cole-h>
For the `zfs destroy` on those snapshots
<sphalerite>
aaah
<sphalerite>
ok, silly idea: try rebooting?
<cole-h>
I did :D
<sphalerite>
aww ok
<sphalerite>
hm, no, I don't think -d would have made a difference — they would still appear in the listing if their destruction were deferred? I think?
<cole-h>
Welp, time to `zfs destroy rpool/system/var@pre-win10-resize`
<sphalerite>
wait
<sphalerite>
what does `zfs get lused rpool/system/var` say?
<sphalerite>
I mean, feel free to destroy it, shouldn't make a difference on that count
<cole-h>
822G
<cole-h>
rpool/system/var logicalused 822G -
<sphalerite>
grmph never mind, what I wanted was "refer". and that's already in your gist
<drakonis>
lmao actually
<drakonis>
i ran out of space
<drakonis>
definitely need more space though
<drakonis>
WILL STILL BUY MORE STORAGE
<bqv>
"space is cheap"... :/
<drakonis>
also
<drakonis>
i have no swap
<drakonis>
for some godforsaken reason i forgot to enable that
<cole-h>
jfkdlsa;jfkl;d
<cole-h>
ZFS is both cool and incredibly arcane.
<bqv>
me neither, but 32G of ram
<drakonis>
i have 16gb of ram
<drakonis>
i'll just set up zram
<cole-h>
Time to ncdu /
<drakonis>
i'll be having none of that swapfile bullshit
<drakonis>
i nuked quakechampions
<drakonis>
it was taking too much space anyways
<drakonis>
i think 25 gigs are used up by symlinks lol
<drakonis>
gdi
<sphalerite>
cole-h: I'm stumped at this point, I've asked #zfsonlinux since my expectation is that `used` of a filesystem ~= `refer` of said filesystem + `used` of its child snapshots
<sphalerite>
(assuming no clones or other fun like that)
<sphalerite>
cole-h: ok, I thoroughly misunderstood
<sphalerite>
cole-h: you probably still have some snapshots with the mistake in them :)
<drakonis>
we should default to zram now
<drakonis>
thanks
<samueldr>
too opinionated, but it should be part of the suggested configuration.nix template
<drakonis>
the new templates system is fairly interesting
__monty__ has quit [Quit: leaving]
<sphalerite>
cole-h: ok, I have an idea. Could you gist `zfs send -nvR rpool/system/var`?
<cole-h>
Error: Unsupported flag with filesystem or bookmark.
<sphalerite>
oops, rpool/system/var@2020-08-10T121500 then
<sphalerite>
cole-h: send from @2020-08-08T100000 to rpool/system/var@2020-08-08T141000 estimated size is 100G so at least the 1410 snapshot is basically eating 100G as well
<sphalerite>
the ones after that may or may not still contain the data, so you might need to rerun the send command and check what's big after that if destroying it doesn't reclaim any space
<bqv>
cole-h: looks easy, ty!
<cole-h>
sphalerite: So you're saying `zfs destroy 1410` and then check free space?
<sphalerite>
cole-h: if you know you don't need the 14:10 state, yep
<sphalerite>
(not that command verbatim of course)
<cole-h>
:P
<sphalerite>
infinisil: just noticed znapzend isn't working right for me, org.znapzend:pre_znap_cmd isn't getting set on one of my datasets. Any ideas off the top of your head?
<infinisil>
Nope
<infinisil>
I'd try torun the prestart script manually
<cole-h>
sphalerite: That just moved the 100G to the next snap.
<sphalerite>
infinisil: aaah, one of my destination names is invalid. So it just gives up on setting anything.
<infinisil>
I see :)
<infinisil>
I do want to switch away from znapzend
<sphalerite>
good call, thanks
<sphalerite>
yeah I've been playing with the idea as well, but… friction
<cole-h>
sphalerite: If I copy the image from the latest snapshot, will that eat up space?
<infinisil>
Yeah
<sphalerite>
cole-h: yeah, copying anything will eat up space
<infinisil>
znapzend at least works for now without any extra effort :)
<cole-h>
Even with `--reflink=auto` (or always)?
<sphalerite>
cole-h: only rollbacks, destroying snapshots, or removing non-snapshotted data will free up space
<sphalerite>
zfs doesn't support reflinks unfortunately
<{^_^}>
openzfs/zfs#405 (by torn5, 8 years ago, open): COW cp (--reflink) support
<cole-h>
Well, eff.
<cole-h>
It's sounding more and more like my only course of action is a rollback........... :(
<sphalerite>
anyway, my suggestion is: destroy all the snapshots except one that you know is good
<sphalerite>
then make a zvol, copy it from the snapshot to that, and remove the snapshot
<cole-h>
Wish I could just destroy the VM image from all snapshots instead...
<sphalerite>
you might be able to do that sort of thing with redacted send. I don't know the details though.
<bqv>
cole-h: do i have to restart for those fonts to take effect, btw?
<sphalerite>
it's also not in 0.8 I think
<cole-h>
djskal;
<cole-h>
bqv: I only had to restart apps like Firefox
<bqv>
hmm
<bqv>
emacs hasn't picked them up, it seems
<bqv>
oh well
<cole-h>
Probably need to restart that too?
<infinisil>
One thing that always annoyed me with znapzend is that you have to really make sure that the SSH connections works, otherwise it just won't work, and that's often a very manual process
<infinisil>
Well that's the case with anything that uses SSH underneath
<bqv>
i did
<bqv>
or, restarted the emacsclient at least
<cole-h>
sphalerite: Well, deleting the pre-win10-resize snap freed up SOME space.
<sphalerite>
infinisil: that makes sense imho, better to fail safe than to send your snapshots to an untrusted location :)
<bqv>
doubt it should require restarting the daemon
<cole-h>
Not enough to copy over the VM image, though... :(
<bqv>
i'll wait for later then, got too much state
<sphalerite>
bqv: yeah the daemon is still the thing making the windows unfortunately, pretty sure you'll need to restart it.
<cole-h>
Because there's a chance fontconfig is loaded by the daemon and not (only) by the client
<bqv>
ah, damn
<infinisil>
And the next step is to write a znapzend module that specifies the SSH connection like this, which allows me to not have to worry about setting it up manually :)
<infinisil>
(or replace znapzend with any other backup tool)
<sphalerite>
ah that's neat
<cole-h>
sphalerite: djklsa and I was supposed to be able to relax today :P
<bqv>
infinisil: omg, i've been wanting something like that for so long
<cole-h>
Stupid human brain
<infinisil>
bqv: Hehe yeah same
<sphalerite>
infinisil: what are you thinking of moving to? sanoid? zrepl?
<infinisil>
sphalerite: Probably sanoid/syncoid for now, but I want to make my own backup tool eventually
<cole-h>
sphalerite: Do you know how I can see the compressed size of the VM image, so I know how much more snapshots I have to delete? :D
<infinisil>
bqv: This is the Nixus module btw, sets up both known_hosts on the source host and authorized_keys on the target host
<bqv>
infinisil: tempting me to see if i can butcher nixus into working for me
<sphalerite>
whoooooa what
<cole-h>
sphalerite: And then `mount` shows all of those mounted on `/.zfs/snapshot/...` instead
<sphalerite>
waaaaaaaat
<infinisil>
bqv: I feel like some other ideas with nixus might work really well with flakes
<infinisil>
And I thought about this too yesterday
<infinisil>
So maybe I'll just take the dive and make it work based on flakes
<bqv>
if you did i'd love you forever
<infinisil>
nixus is experimental, flakes is experimental, that's a good fit
<bqv>
lol
<cole-h>
sphalerite: Maybe because of nixos...
<cole-h>
Y'know, with all the symlinks in /var
<sphalerite>
huh?
<sphalerite>
is anything along the path /var/lib/libvirt/images a symlink for you?
<bqv>
infinisil: i actually think nixus is the only one out of the candidates that could properly integrate with flakes, because it's almost entirely nix-based
<cole-h>
Don't think so sphalerite
<bqv>
what are your design thoughts?
<cole-h>
sphalerite: Actually, `/var/lib/alertmanager` is a symlink...
<sphalerite>
yeah but that shouldn't be relevant..?
<sphalerite>
does the du command work?
<infinisil>
bqv: Wait let's go to #nixus
<infinisil>
Quite full in here
<bqv>
i'm pleased emacs makes that link work for me
<cole-h>
sphalerite: No. `du: cannot access '/var/.zfs/snapshot/*/lib/libvirt/images/windows10.img': No such file or directory`
<sphalerite>
cole-h: did I get part of the path wrong?
<sphalerite>
oh right .img instead of .raw
<cole-h>
Oh, shoud be .raw
<sphalerite>
at the very least
<sphalerite>
infinisil: samueldr: can we have logs for #nixus? :)
<cole-h>
sphalerite: Same output, though.
<samueldr>
would be possible
<sphalerite>
cole-h: with the too-many-symlinks?
<cole-h>
No, doesn't exist
<sphalerite>
cole-h: maybe another bit of the path is wrong?
<sphalerite>
how about ls /var/.zfs/snapshot/*/lib/libvirt/images/ ?
<sphalerite>
Oh or maybe you need to do the globbing as root because permissions?
<cole-h>
Because accessing /var/.zfs/snapshot/* mounts it to /.zfs for whatever reason
<cole-h>
(idefk)
<sphalerite>
I don't even
<sphalerite>
<3 samueldr
<{^_^}>
samueldr's karma got increased to 254.99999999999986
<sphalerite>
cole-h: so that should tell you how much space each version takes, with the caveat that this space is shared between them, depending on Things
<infinisil>
Oh I might need to manually correct {^_^} a bit, otherwise the float drift gets too big eventually!
<infinisil>
I really wish there was a way to not have to store karma as a float!
<cole-h>
sphalerite: Thanks for your help. Somehow /var/.zfs/snapshot/* gives ELOOP lol idek
<samueldr>
I knew about all that, but the way it was stated confused me to a higher plane of existence
<samueldr>
where I couldn't discern whether an haskell-based irc bot really needed to store values as floats for some awful reason
<sphalerite>
you mean the code isn't obvious??
<sphalerite>
:D
<infinisil>
The code is awful lol
<samueldr>
do I look like I know what a monad is? all I wanted was to get a got dang line of text on this got dang irc
<cole-h>
sphalerite: If you were me, what would you use to copy the image from snapshot to its rightful place? Plain cp? rsync for viewing progress? Something else?
<cole-h>
(Finally have 313G of free space)
* samueldr
wonders how badly explorer.exe through wine would fare
<sphalerite>
cole-h: probably pv, since rsync doesn't like block devices. If you're moving to a zvol.
<cole-h>
sphalerite: Mmm. Is a zvol just `zfs create`? I've always been unclear on that term.
<sphalerite>
so pv /.zfs/snapshot/…/windows-10.raw > /dev/zvol/rpool/windows-10
<cole-h>
sphalerite: Should $size be what `du -h` the image returns?
<sphalerite>
no, it should what du -h --apparent-size on the image returns
<sphalerite>
so the number of gigabytes you want available to dinwos
<sphalerite>
wow that was an impressive typo
<sphalerite>
e.g. 100G
<sphalerite>
what's nice about having it as a block device is that it allows easy mounting of it in linux as well (though don't do that while the VM is running) e.g. mount -o ro /dev/zvol/rpool/windows-10-part2 /mnt
<cole-h>
OK, du says 302G, so that's what I'll give it. Can this be grown/shrunk as needed?
<sphalerite>
yes
<sphalerite>
using zfs set volsize
<sphalerite>
careful, it will truncate the image ruthlessly if you make it smaller :)
<sphalerite>
(though existing snapshots won't be affected)
<cole-h>
Hm. I can use qemu-img to shrink it based on the file, e.g. `qemu-img resize -f raw $imgpath -$sizeG`. Would I just set $imgpath to /dev/zvol/rpool/windows-10...?
<sphalerite>
pretty sure qemu-img will just truncate a file
<dottedmag>
Folks, I have sent several questions to #nixos over several days, all strictly topical and got absolutely zero response, now I wonder if I am filtered out there somehow. How do I know?
<cole-h>
Well, it seems nicer than letting zfs do it :P
<sphalerite>
so you'd use zfs set volsize=$size rpool/windows-10 _instead_ of that
<cole-h>
dottedmag: Maybe nobody knows the answer. Maybe you're asking at the wrong times. I can certainly see your messages (here, at least).
<sphalerite>
^
<dottedmag>
OK, could you check you see my msgs in #nixos too?
<cole-h>
Or maybe you're asking the wrong questions.
<cole-h>
I've got my hands full at the moment :)
<cole-h>
dottedmag: I misread your question; I can see your messages over there, too.
<dottedmag>
thanks.
<cole-h>
sphalerite: I assume I just need to point libvirt to the /dev/zvol path afterwards, right?
<sphalerite>
cole-h: yup. It supports zfs pools as storage somehow as well, if you're using virt-manager you can set it up on the connection properties iirc
tazjin has quit [Quit: Byeee]
andi- has quit [Ping timeout: 260 seconds]
tazjin has joined #nixos-chat
<cole-h>
Ah nice, even more free space: 36092 store paths deleted, 45470.18 MiB freed
<cole-h>
sphalerite: Wait
<cole-h>
Maybe the problem with ELOOP is that /usr/bin/env is a symlink to coreutils and it doesn't like that?
<sphalerite>
uuuuuuuh
<sphalerite>
what's /usr/bin/env got to do with it?
<cole-h>
sphalerite: And that symlink to coreutils/bin/env is a symlink to coreutils itself
<sphalerite>
hm, ok. But I don't think that should cause such an issue.
<cole-h>
Hm, no. I can `ls /home/.zfs/snapshot/snap` just fine..
<cole-h>
However, `ls`ing those DOES make an entry under `mount`!
<cole-h>
sphalerite: Can you try the same? Just `ls` (or otherwise interact with) any of your snapshots and see if it creates a mount entry?
<sphalerite>
yes, that's normal
<cole-h>
oh
<cole-h>
sik
<cole-h>
But for some reason /var/.zfs mounts its snapshot to /.zfs....
<cole-h>
Brain exploded
<sphalerite>
that does seem like a bug
kalbasit has joined #nixos-chat
<cole-h>
sphalerite: Do you think a zvol will have better performance than a file? Or will it not make a difference
<cole-h>
I've seen reduced performance ever since switching to NixOS + ZFS -- I imagine it's due to the fact I enabled both encryption (and -ccm at that) and compression, but I don't have anything to back that up other than "a feeling".
<sphalerite>
not sure
<sphalerite>
compression should result in _better_ performance unless you selected an expensive algorithm
<cole-h>
sphalerite: I only did `compresion=on`, so it uses whatever the default is, I guess
<samueldr>
it would be hilarious if it was `cat`
<cole-h>
heh
<sphalerite>
yeah that's sensible.
<sphalerite>
for VM images it might make sense to disable compression or only use zle, but I haven't tested/benchmarkes this myself so it's only hearsay from me.
andi- has joined #nixos-chat
<sphalerite>
samueldr: md5sum!
dadada_ has quit [Ping timeout: 256 seconds]
<samueldr>
isn't that somewhat undeterministic?
<samueldr>
multiple inputs for the same output (or vice-versa)
<sphalerite>
who cares? It's got a great compression ratio
<samueldr>
might as well use jpeg encoding
<samueldr>
or h264 and have your gpu do the work
<sphalerite>
oooooh that's a good shout
<samueldr>
(or whatever the actual frame compression method is used within h264)
parsley936 has quit [Remote host closed the connection]
* sphalerite
is tempted to start building his own kernels with localyesconfig after seeing how easy it is
<sphalerite>
help
dadada_ has joined #nixos-chat
<cole-h>
Why and what does that do?
<samueldr>
cole-h: yes
<samueldr>
but locally
<cole-h>
lol
<sphalerite>
cole-h: builds all the modules loaded in the running system into the kernel
<cole-h>
Does that work in the sandbox?
<sphalerite>
I'm not sure
<sphalerite>
I'd generate the config outside the sandbox though