leeola has quit [Quit: Connection closed for inactivity]
subtraktion has quit [Remote host closed the connection]
cemerick has joined #ipfs
spacebar_ has quit [Quit: spacebar_ pressed ESC]
galois_dmz has quit []
ebarch has joined #ipfs
ashark has joined #ipfs
ashark has quit [Ping timeout: 264 seconds]
maxlath has quit [Quit: maxlath]
galois_dmz has joined #ipfs
tilgovi has quit [Remote host closed the connection]
galois_d_ has joined #ipfs
galois_d_ has quit [Client Quit]
galois_d_ has joined #ipfs
galois_dmz has quit [Ping timeout: 256 seconds]
tilgovi has joined #ipfs
galois_dmz has joined #ipfs
cemerick has quit [Ping timeout: 246 seconds]
cemerick has joined #ipfs
galois_d_ has quit [Ping timeout: 256 seconds]
galois_dmz has quit [Ping timeout: 256 seconds]
bwn has quit [Ping timeout: 260 seconds]
bwn has joined #ipfs
anewuser has quit [Ping timeout: 240 seconds]
<whyrusleeping>
kevina: i'm around
realisation has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<whyrusleeping>
kevina: the approach looks good to me
<whyrusleeping>
though it doesnt build :P
<kevina>
whyrusleeping: one second let be back out a commit...
tmg has joined #ipfs
cemerick has quit [Ping timeout: 246 seconds]
<kevina>
whyrusleeping: ok. the p.r. is ready for review
<whyrusleeping>
kevina: jenkins still saying it doesnt build...
<kevina>
whyrusleeping: it build on my machine, checking to see what circleci says
<kevina>
whyrusleeping: ci/circleci passes
Dunkhan has quit [Read error: Connection reset by peer]
<kevina>
whyrusleeping: okay if failed after a merge will rebase and fix
<lemmi>
whyrusleeping: so leveldb seems to have a higher constant latency but doesn't grow out of hand. but i just saw leveldb has an option for in memory storage, too.
<whyrusleeping>
oooh, interesting. Let me know how that one goes
<whyrusleeping>
you should have write access there
<kevina>
whyrusleeping: wasn't the idea that they will be moved automatically?
realisation has joined #ipfs
<whyrusleeping>
kevina: no, i think Kubuxu was just doing it by hand
<Kubuxu>
there is script in repo that does it if branch is older than given time
<Kubuxu>
I can put in on cron somewhere
<Kubuxu>
but so far I was planning to run it ever so often by hand
<Kubuxu>
feel free to put it in there manually either way
<kevina>
Kubuxu, whyrusleeping: okay will do manually in a bit, need to concentrate right now on fixing the Cidv1 flag p.r.
<whyrusleeping>
kevina: sounds good, thanks :)
realisation has quit [Max SendQ exceeded]
melvster has quit [Ping timeout: 268 seconds]
ianopolous has joined #ipfs
<lemmi>
whyrusleeping: memstorage seems to work very well. i can't see any real memory difference to using files, but latency is easily an order of magnitude better
<lemmi>
(and it doesn't hurt the ssd so much)
<whyrusleeping>
lemmi: ah nice, see how it works for a few hours
<whyrusleeping>
then send me the patch :D
<lemmi>
whyrusleeping: i let it cook for for the day and report back
<lemmi>
whyrusleeping: :\ i have the suspicion that the latency measurement is slightly misleading. i got a jump around 6000 entries from ~1ms to over 230ms and now constantly decaying. maybe keeping track of the variance, identifying largy outliers and keeping track of them separately might help.
<whyrusleeping>
lemmi: hrm.... good point
<whyrusleeping>
i'm just using a direct average
<lemmi>
a couple moving averages could also help
<lemmi>
but if a single transaction takes seconds and the rest is fine, one probably wants to know about that.
<whyrusleeping>
lemmi: if youre feeling adventurous, i definitely accept PRs :)
<lemmi>
yeah.. i have a bachelor thesis to finish by the end of next week, so don't expect miracles until then :)
<whyrusleeping>
haha, no worries
<whyrusleeping>
If i find time i'll get to it too
<whyrusleeping>
its now a race to see who finds free time first
<kevina>
whyrusleeping: actually where I was not that cold in the winter, it was fairly south and by the sea. In the winter the temperature was a constant 30 deg (-1 cel)
<kevina>
by constant I mean there was no high or low during the day. If you look at a temp graph you would not be able to tell when the sun was up
<kevina>
the summers where really nice, you could still see at midnight
<whyrusleeping>
kevina: oh yeah, by the sea is always decent
<whyrusleeping>
nice :)
<whyrusleeping>
Kubuxu: okay fine, you have good dirt
<Kubuxu>
lol
<whyrusleeping>
we have AR15s, Beef, The rocky mountains, and san francisco
<Kubuxu>
if it comes in packet with high incarceration rate, wealth gap and no health system (nope what you currently have is no a system), you can have it
<Kubuxu>
I like this dirt
<whyrusleeping>
We have a great healthcare system. You just have to have a job
<Kubuxu>
old, "to not be poor, don't be poor"
<whyrusleeping>
Yeap
<whyrusleeping>
welcome to capitalist america
<Kubuxu>
our Pope said long time ago: "A society will be judged on the basis of how it treats its weakest members"
IRCFrEAK has joined #ipfs
IRCFrEAK has left #ipfs [#ipfs]
<whyrusleeping>
Kubuxu: hrm....
<whyrusleeping>
I don't know about that
<whyrusleeping>
what with capitalism and all
_IRCFrEAK has joined #ipfs
_IRCFrEAK has left #ipfs [#ipfs]
captain_morgan has quit [Remote host closed the connection]
<whyrusleeping>
i'll bet youre hitting swap or something
<lemmi>
nope
<lemmi>
RES is 120mb
<lemmi>
no problem there
<whyrusleeping>
I've had one running with normal leveldb for 27 hours now, 180k records, and 800us latency
<whyrusleeping>
why would it randomly spike so high?
<lemmi>
no idea. it also seems to stay at 30700 records
<lemmi>
no movement at all
<lemmi>
oho... and suddenly a whole bunch and the latency shoots up, too
<lemmi>
peaked at ~ 30s latency
<whyrusleeping>
o.o
<lemmi>
and cpu is down to 10%
<lemmi>
was over 100% for a while
<whyrusleeping>
whoa
<lemmi>
something's wonky
<whyrusleeping>
Yeah, agreed
<whyrusleeping>
I'm running a machine with 10 nodes and it sitse around 7% cpu
shoragan has quit [Ping timeout: 264 seconds]
mbags has quit [Ping timeout: 260 seconds]
<lemmi>
that kind of latency lets me believe that some network connection is maybe in a bad state, the records aren't concurrently accessable and then it hangs. the program doesn't have any disk access, so that's unlikey. i'd be surprised if it's leveldb. ulimits are changed to high values for open fds.
<lemmi>
don't have much insight into what's actuall going on, just guessing
<whyrusleeping>
Yeah... I'm gonna push an update that allows you to expose a pprof profiling server
<whyrusleeping>
so you can use go tool pprof to get cpu and memory profiles
<lemmi>
yeah. it was eating a whole cpu while it was hanging
<whyrusleeping>
pushed
<whyrusleeping>
if you run it with -pprof-port=3000 it will expose pprof info on port 3000