<makefu>
colemickens: for me it simply became part of the nixos system configuration. there is no reason why to remove it. at work i am using home-manager to get at least a bit of sane system management on a centOS host
<ashkitten>
they're so confusing and i10e (inaccessible)
<infinisil>
> numeronym "ashkitten"
<{^_^}>
"a7n"
<ashkitten>
i didn't realize what r13y was for like
<ashkitten>
an entire year
<gchristensen>
that is impressive, I think I only registered it 51 weeks ago =)
<cransom>
fun story from a different life. a friend was at a matrix (yes, the movie) themed company. their windows domain controller was called neo. the boss decided one day that it should really be named ne0 and didn't tell anyone. it took down the domain for a day until someone figured it out.
<gchristensen>
I really just don't know about smart stuff enough, I guess
<tilpner>
gchristensen: Hmm, that's a lot more data than I wanted to handle :)
<gchristensen>
tilpner: yeah, the thing about SMART, to me, is it gives me a bunch of data I don't know what to do with.
<gchristensen>
and really whaht I want is just a "hey something is fishy, maybe you should buy a new disk."
<tilpner>
gchristensen: The plan I have/had till now, was to modify the smartd module to run a script instead of sending an email, and then pass this notification to prometheus via textfiles, but it's not really how prometheus is meant to be used
<gchristensen>
it does expose smartmon_device_smart_healthy but I don't know if that would go from 1 to 0 before things were severely unhealthy
<tilpner>
That would get me out of defining what an emergency looks like though
<gchristensen>
true, setting up some alerts on that would be good I suppose
<gchristensen>
the thing about having smartd run the script is the script will only run when things are bad, and it is hard to be sure it will actually work
<gchristensen>
running it every few minutes, you get a promise that it is still working (assuming you export metrics about when the data is from, and alert on that data becoming old -- an important step since textfile exporters can fail and the data is still being loaded)
<tilpner>
Yes, you make good points. Thank you for the scripts, I'll try to understand what it does :)
<samueldr>
there's this one thing I haven't really been able to find much about grafana/prometheus
<gchristensen>
tilpner: yep! if you take a further look, would you mind taking over my PR and try to get it merged? :P
<samueldr>
let's say I want to collect data on e.g. my laptop, but have it on a centralized dashboard, and that dashboard is only locally accessible when I'm home
<samueldr>
is there a good strategy?
<samueldr>
does it just implicitly happen?
<gchristensen>
samueldr: prometheus scrapes all the metric exporters a few times a minute (configurable.) you'd need to make it so your prometheus can always scrape your laptop no matter where you are. I do that with a wireguard tunnel
<samueldr>
hm, bummer
<gchristensen>
yeah?
<samueldr>
I would have preferred that the device keeps the data locally, and push it to be graphed
<samueldr>
I should really look in that 19.09 regression in the munin module
<samueldr>
or uh, just see if it has been solved already
<gchristensen>
samueldr: prometheus strongly prefers the pull model. you might prefer influxdb which uses a push model
<tilpner>
gchristensen: I will make no such promises c.c
<gchristensen>
tilpner: sure :)
<samueldr>
I only said prometheus since it seems used across the board
<samueldr>
but in reality I want to graph data prettily
<gchristensen>
samueldr: one reason prometheus prefers pull is a failed scrape is usually an indication of a problem, but it is more tricky to instrument lack of data being pushed in
<samueldr>
I don't think I actually need prometheus
<samueldr>
I think I need whatever munin does, but graphed in grafana
<gchristensen>
yeah but also transferrable skills and advantages of every other infra nerd writing exporters for things like their cable modem https://github.com/ipstatic/surfboard_exporter
<samueldr>
my need is not alerting or fancyness, but really pretty graph of my usage
<tilpner>
I don't actually use grafana anymore, but the (very annoying) alerts from alertmanager are useful
<samueldr>
mostly for "oh, look, I finally used about a third of the RAM of my woefully overspecced builder"
<gchristensen>
yeah, prometheus doesn't really do alerting :P
<tilpner>
Just the "there is a failed systemd service" alert even
<samueldr>
I only want pretty graphs that a raspberry pi can show later :)
<gchristensen>
anyway, I'm not going to push you to prom
<gchristensen>
but moving with the herd sometimes is useful
<tilpner>
samueldr: Does netdata do what you want?
<tilpner>
Lots of pretty graphs, not useful for much else
<gchristensen>
samueldr: also I do have a self-interest in you using prom, actually, which is you can help with the infra =)
<tilpner>
And just a services.netdata.enable away
<samueldr>
gchristensen: I'm not dismissing prometheuse outright, simply that it's not the use case I have for right now
<gchristensen>
not inerested in a wg tunnel?
<samueldr>
not necessarily for or against, but not interested in working that way for that silly need, since what I want is for sillyness and not for real use
<gchristensen>
ah
<samueldr>
it's more of an informal data aggregation than actual logging for my personal use machines
<samueldr>
on serverey machines I would use the real deal though and will look at some point
<gchristensen>
I've found it quite fun to have a prom. it is easy to write a little data exporter and graph data without much work.
<gchristensen>
I have a SDR recording my gas and electricity usage every minute
<tilpner>
Software-defined radio?
<samueldr>
it's really more about the use case of just recording the data of my personal usage machines and showing it for fun
<gchristensen>
yeah
<samueldr>
gchristensen: here's a page where scrolling is bad in firefox https://www.netdata.cloud/ (ref discussion a couple days ago)
<jtojnar>
wut, I am getting some spam from "github musicians"
<samueldr>
only when the animation is visible, just like I observed on "edgy" blogs with "meme gifs" strewn across
<samueldr>
this strips language-specific folders from the output
<samueldr>
mostly
<samueldr>
you can spot a couple packaging bugs in there :)
<samueldr>
just checked with rox, which I have installed globally
<samueldr>
No manual entry for rox
<tilpner>
Huh, vim can edit .gz files directly
<tilpner>
Not that I would edit .gz manpages in the store
<samueldr>
nix-locate -1r 'share/man/[^/]+.gz' # for a list of hopefully trivial to fix packages
__monty__ has quit [Quit: leaving]
<samueldr>
gosh nix-locate is amazing at getting a holistic view of nixpkgs build artifacts <3 bennofs
<{^_^}>
bennofs's karma got increased to 1
<infinisil>
We do have this neat `installManPage` function since recently
<samueldr>
(aw, I guess the [m] variant of bennofs hoards all the karma
<infinisil>
Oh! That's like the perfect usecase for a union-find datastructure
<samueldr>
of a what now?
<infinisil>
Managing which nicks belong together
<samueldr>
oh, I thought it was about man pages
* infinisil
should've indicated that
<samueldr>
I really like using nix-locate and good regexes to find things like what package provides a particular theme
<samueldr>
nix-locate -r 'share/themes/[^/]+$'
<infinisil>
Yeah nix-locate is awesome, bennofs[m]++
<{^_^}>
bennofs[m]'s karma got increased to 5
Jackneill has joined #nixos-chat
<samueldr>
throw in a --top-level to my previous command for better results
Jackneill has quit [Remote host closed the connection]
waleee-cl has quit [Quit: Connection closed for inactivity]
<colemickens>
someone recommending a forked python lib renamed from 'x' to 'python-x' and I'm confused on how this is meant to be helpful
<colemickens>
reminds me of when there are rust-x and x-rs or when a crate repo name doesn't match its published names. why not just try to not make things confusing for lib consumers? Classic computing problems I guess