Leo_Verto: ah yes, just saw that, will sort it out tomorrow
LordSputnik has quit
Freso
Anyone here know how to specify negative niceness via schedtool?
JesseW has quit
JesseW joined the channel
JesseW has quit
JesseW joined the channel
_iwc joined the channel
_iwc has left the channel
_iwc joined the channel
_iwc has left the channel
_iwc joined the channel
_iwc has left the channel
_iwc joined the channel
_iwc has left the channel
_iwc joined the channel
_iwc has left the channel
_iwc joined the channel
_iwc has left the channel
_iwc joined the channel
_iwc has left the channel
reosarevok joined the channel
reosarevok has quit
reosarevok joined the channel
Gentlecat has quit
yeeeargh joined the channel
UmkaDK joined the channel
samphippen joined the channel
CatCat has quit
CatCat joined the channel
zas joined the channel
JonnyJD joined the channel
Nyanko-sensei joined the channel
D4RK-PH0ENiX has quit
yeeeargh has quit
samphippen has quit
Mineo
listenbrainz seems to be returning a lot of 502s
Lotheric has quit
samphippen joined the channel
Nyanko-sensei has quit
D4RK-PH0ENiX joined the channel
JesseW has quit
Nyanko-sensei joined the channel
samphippen has quit
D4RK-PH0ENiX has quit
Freso
alastairp: ^ (no ruaok awake yet it seems)
kahu_ has quit
Nyanko-sensei has quit
D4RK-PH0ENiX joined the channel
D4RK-PH0ENiX has quit
D4RK-PH0ENiX joined the channel
alastairp
I just restarted it. we’re still having some problems with it
darwin
what is "it" in this casE?
an app server?
Nyanko-sensei joined the channel
ariscop has quit
D4RK-PH0ENiX has quit
alastairp
yes
flask + uwsgi
we’re having problems where some workers seem to get stuck after about a day on high cpu, and uwsgi stops sending requests to them
Nyanko-sensei has quit
D4RK-PH0ENiX joined the channel
darwin
workers live that long?
D4RK-PH0ENiX has quit
Nyanko-sensei joined the channel
alastairp
web workers. Yes, they’re supposed to be long running, and nginx passes requests to uwsgi, which is supposed to hand it off to a worker
russss
alastairp: can you tell uwsgi to restart them after serving n requests?
darwin
I'm pretty sure that's what we do at my work.
like, I would not expect them to run "about a day"
ruaok joined the channel
(but I am "just the DBA" hehe)
D4RK-PH0ENiX joined the channel
Nyanko-sensei has quit
D4RK-PH0ENiX has quit
D4RK-PH0ENiX joined the channel
Nyanko-sensei joined the channel
Nyanko-sensei has quit
Nyanko-sensei joined the channel
D4RK-PH0ENiX has quit
samphippen joined the channel
ariscop joined the channel
samphippen has quit
samphippen joined the channel
alastairp
russss: it has so many options, I'm sure it's a possibility
though we've never had this problem on acousticbrainz (definitely less submissions though)
russss
might be worth doing that as a first possibility. It could be there's a resource leak somewhere
but finding those is hard
the other alternative is that occasionally something is looping for ever
reosarevok has quit
reosarevok joined the channel
alastairp
yeah, those are our thoughts too. and we also know that finding them is hard
one thing I wanted to do was attach strace to a high cpu processes
just to see what it's doing
russss
there's a high probability it'll show either nothing, a load of select()s on the epoll socket, or a load of futex_wait. None of which is helpful :)
but worth a try. (also ltrace)
alastairp
ah, never seen ltrace
oh right, strace is system calls, ltrace is library calls. neat
ruaok has quit
russss
still waiting on a good dtrace-alike for Linux which works out of the box. Apparently Systemtap may have userspace tracing support for python now, but I need to spend some time looking at it.
Laverne
sysdig might be interesting to look at as well
russss
sysdig is cool but it's basically just a new frontend to existing tools rather than a new datasource.
samphippen has quit
Mineo
were there any problems with the caa-indexer in the last month? I had to perform some edits on https://musicbrainz.org/release/0df9fe6e-d374-4... to make it return the front type in the json file correctly