wyatt has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
maxlath has joined #ipfs
gmcabrita has joined #ipfs
hoenir has joined #ipfs
ylp1 has joined #ipfs
ylp has quit [Quit: Leaving.]
<musicmatze>
r0kk3rz: ah, okay... so I guess I'm doing something wrong. (I'm just trying to write a little micro-blogging thingy, with a commandline interface, never did smth with js/node/whatever ... so that'll be a really new experience for me :D)
<musicmatze>
I have a question about running nodes: I run two nodes in my universities cloud, and "bmon" tells me that there is about ~20KiB/s outgoing traffic - I assume that my nodes do not share any pinned content right now (although I pinned a few things), so I think this is basic communication for the IPFS network to work, am I right?
<r0kk3rz>
musicmatze: its the DHT chatter
Foxcool has joined #ipfs
mazeinmaze_ has quit [Ping timeout: 264 seconds]
<musicmatze>
r0kk3rz: yup, then my assumptions were correct :-) Awesome! So if someone needs the linux kernel repository, nixpkgs, nix itself, git,... :-) I have them pinned on two machines with big bandwidth :-)
ianopolous_ has joined #ipfs
ianopolous has quit [Read error: Connection reset by peer]
<musicmatze>
r0kk3rz: To come back to what you said... ipfs-js is a node on its own ... so it would be kinda senseless to write a one-shot commandline application with it, right? Because if the app terminates, the node is gone and so is my data (if I didn't pin it elsewere in between)
<musicmatze>
so I guess I have to use ipfs-js-api or even write my application in either go or (as far as I know there is a library for) c++?
<r0kk3rz>
well its just a wrapper for the http API
<r0kk3rz>
so as long as someone updates it, then it doesnt matter whether its 'official' or not
sirdancealot has quit [Ping timeout: 240 seconds]
sirdancealot has joined #ipfs
antfoo_ has joined #ipfs
antfoo has quit [Ping timeout: 240 seconds]
antfoo_ has quit [Ping timeout: 240 seconds]
antfoo has joined #ipfs
shrimpx_ has quit [Ping timeout: 240 seconds]
espadrine has joined #ipfs
f[x] has quit [Ping timeout: 256 seconds]
Kamilion has quit [Read error: Connection reset by peer]
hoenir has quit [Ping timeout: 260 seconds]
azahi has joined #ipfs
hoenir has joined #ipfs
Foxcool_ has joined #ipfs
Foxcool has quit [Ping timeout: 268 seconds]
shrimpx has joined #ipfs
Kamilion has joined #ipfs
ygrek has joined #ipfs
engdesart has quit [Quit: yes]
jkilpatr has joined #ipfs
ygrek has quit [Ping timeout: 268 seconds]
Rob has joined #ipfs
Rob is now known as Guest5062
<Guest5062>
Hi Guys can anyone help with finding out the cost of using IPFS
<Guest5062>
in terms of different metric such as time, space
robinkatoch has joined #ipfs
Foxcool_ has quit [Ping timeout: 260 seconds]
hoenir has quit [Remote host closed the connection]
dignifiedquire has joined #ipfs
cxl000 has joined #ipfs
ygrek has joined #ipfs
antfoo has quit [Remote host closed the connection]
antfoo has joined #ipfs
dimitarvp has joined #ipfs
m3s has quit [Remote host closed the connection]
Foxcool has joined #ipfs
Foxcool has quit [Ping timeout: 260 seconds]
ygrek has quit [Remote host closed the connection]
ygrek has joined #ipfs
PrinceOfPeeves has joined #ipfs
ygrek has quit [Ping timeout: 260 seconds]
s_kunk has joined #ipfs
s_kunk has joined #ipfs
s_kunk has quit [Changing host]
ulrichard has joined #ipfs
Mitar has quit [Ping timeout: 256 seconds]
<musicmatze>
can someone explain to me why the go ipfs impl spawns hundrets of threads?
ianopolous_ has quit [Remote host closed the connection]
<musicmatze>
I just used the webui for a bit (nothing fancy, just clicking around) and I had 100 threads only for IPFS; the whole thing running with ~50-150% CPU usage ... doing what?
hoenir has joined #ipfs
Oatmeal has joined #ipfs
NiXZe has quit [Ping timeout: 240 seconds]
<lemmi>
IIRC go spawns os threads for potentially blocking syscalls. ipfs is quite busy with joining the network, lots of connections, little bit of disk usage, dht and all.
<lemmi>
so most threads should more or less only sit there and idle.
NiXZe has joined #ipfs
NiXZe is now known as Guest75024
ianopolous_ has joined #ipfs
espadrine_ has joined #ipfs
espadrine has quit [Ping timeout: 258 seconds]
Foxcool has joined #ipfs
saintromuald has quit [Remote host closed the connection]
citizenErased has quit [Quit: Leaving]
jkilpatr has quit [Ping timeout: 246 seconds]
shizy has joined #ipfs
Captain_Beezay is now known as \-__-_____----_
<whyrusleeping>
Guest5062: what exactly are you looking for?
<whyrusleeping>
musicmatze: what OS?
hoenir has quit [Ping timeout: 246 seconds]
tmg has quit [Ping timeout: 268 seconds]
ashark has joined #ipfs
leeola has joined #ipfs
sirdancealot has quit [Ping timeout: 256 seconds]
Reventlov has quit [Quit: WeeChat 1.7]
Reventlov has joined #ipfs
hoenir has joined #ipfs
<lgierth>
PSA: all hands calls in 1h15m -- we'll be evaluating how Q1 went
<musicmatze>
whyrusleeping: NixOS linux
<musicmatze>
I even managed to get one ipfs node on one of my VMs up to ~2,4k threads
<lgierth>
about the all hands calls, we're not set up for streaming, so dial into the zoom meeting (TBD) to listen in or participate
<whyrusleeping>
musicmatze: hrm... thats not entirely weird, but 100 is on the high end of what i observe
<whyrusleeping>
the primary issue is (as lemmi said) that go will spawn a new thread whenever it thinks some syscall may block
<whyrusleeping>
the syscall in question here is tcp socket operations
<whyrusleeping>
We're hoping to switch to udp based transports like QUIC soon
<whyrusleeping>
which should address a lot of these resource consumption issues
<musicmatze>
I would say "Lets rewrite it in $randomotherlanguage" ... but nope :-) that would end in endless bikeshedding :-)
<whyrusleeping>
musicmatze: heh, bikeshedding and wouldnt actually solve any problems
<whyrusleeping>
this isnt a language issue
<musicmatze>
ah, okay... Wrong assumption on my end.
<whyrusleeping>
done in another language, we would have to manually implement something similar to what go does for us
<whyrusleeping>
the main problem here is that we're giving go so many things that can block
<whyrusleeping>
and its freaking out like "holy crap, what are all these blocking things? new thread, new thread, and YOU get a new thread"
<Kubuxu>
and it never frees threads
<musicmatze>
Kubuxu: that's the real issue I guess
m3s has joined #ipfs
m3s has joined #ipfs
m3s has quit [Changing host]
Guest5062 has quit [Quit: Page closed]
wmoh has joined #ipfs
jonnycrunch has joined #ipfs
<whyrusleeping>
heh
<whyrusleeping>
well
<whyrusleeping>
the *really real* issue here may actually be that we're circumventing a lot of go's network stack
<whyrusleeping>
because they don't yet support SO_REUSEPORT options on tcp listeners