which returns artist credit for the recording but how to get the artist credits for induvidual releases
reosarevok
amCap1712: add artist-credits to the inc IIRC
amCap1712
thanks reosarevok
it worked
so that means artist parameter queries artist details for the entity and artist credit for associated entities?
Lotheric has quit
Lotheric joined the channel
Nyanko-sensei has quit
D4RK-PH0ENiX joined the channel
Freso
amCap1712: And if possible, you should use https:// always, please. :)
amCap1712
Freso: I use https by default but I have an extension to automatially pretty print json in browser and i don't know but it behaves unusually with https. Thats why the http link.
hmmm, actually not sure what happened on lemmy, did someone restart any container on it recently ?
ruaok
recently?
no
but I have no idea why that proxy container keep flapping.
well, the container is ilde and doing nothing. not getting any requests at all.
but the service blinks on and off all the time.
zas
it seems related to the error reo got, which isn't good, because it applies to all pg bouncers connections
ruaok
I think with BB on board, we may need to increase the number of active connections we allow to PG.
zas
something changed on lemmy at the same time, i thought it was a container restart, but apparently not
ruaok
if it was, I didn't do it.
zas
ruaok: 400 simultaneous connections is more than enough, but i think something was actually using too many connections during a short instant, it's now back to normal
ruaok
actually, if you could look at the proxy-api container issue, I'd love you for it. it doesn't appear to be a problem with the service, but with traffic getting to the service.
zas
the usual number of connections is around 100
ruaok
zas: pg bouncer is keeping a lot of connections open for pooling.
zas
yes, but values are ok, and safe imho, increasing this number doesn't fix the problem we just had. In fact, it is much safer to have this number not too high. The real question is why, suddenly, the number of connections increased that much (>800) = which service actually did this shit, i suspect lb stuff, because it seems to me everything points at lemmy, but i can be wrong ofc
ruaok: where is running this proxy-api container ??
ruaok
lemmy
zas
i see none with this name
ruaok
listenbrainz-api-compat-prod
zas
ah, ok, let me see
did you look at its log ?
because imho it's the cause of the problem, it was spawning/killing workers at the same time the reo's error happened
and I dont understand it either. for some reason a consul value that we defined just doesn't show up.
then on the next restart it does.
iliekcomputers: can you please rebuild the compat api container with the latest brainzutils image and redeploy it?
that might solve the problem -- upgrade all the stuff to the latest and greatest.
CatQuest
in my experience upgrading shit tends to break shit but ¯\_(ツ)_/¯
yokel joined the channel
D4RK-PH0ENiX has quit
zas
ruaok: i think i found the issue, uwsgi doesn't check for config files to exist before starting, so it is possible consul-template didn't generate the config file yet, while uwsgi is starting
also, processes is set to 100, but i see only 50 workers (dunno why), threads is not set (so it defaults to 1), but each process will create 3 connections to pg on bowie (it uses 3 dbs), and 1 to pg on queen, it means 300 connections, which is likely to exhaust all connections pretty fast on high load. Not sure if i understand this uwsgi stuff well, but since each process is totally separated from others, pg connections aren't
shared imho
bitmap, yvanzo: do we have a mean to know how many connections are made to pg on bowie from lemmy ?
D4RK-PH0ENiX joined the channel
yvanzo
lemmy is listenbrainz+messybrainz, right? 9+10 atm
ferbncode
spellew: The timeline looks good. Let me know if you have something to discuss before the coding period starts. Also, There is an issue up in CB related to the `musicbrainz_db`. You'd probably want to work on it before you start your actual gsoc project. It would be great if you could take a look at it (https://sentry.metabrainz.org/share/issue/a71fd...).
CatQuest
meow `😺`
pristine__
ruaok: Good evening:)
presently, all the three scripts are dependent, so for instance if we need to check something in recommend.py (script responsible to generate recommendations) we shall wait till create_dataframes.py and train_models.py run. Should we make them independent?
I mean storing the result of every script in hdfs so that the script that follows can retrieve the required data from hdfs and not wait for other scripts to finish.
Cyna
yvanzo, bitmap : the code for aliases component is in .tt right ?. I found it in /root/entity/alias/edit.tt
so I guess my first task is to convert it to React right ?
Lotheric has quit
Lotheric_ joined the channel
Lotheric_ is now known as Lotheric
SothoTalKer_ joined the channel
reosarevok
Cyna: you don't need to convert that, that's the actual form for editing aliases
I mean, eventually we should convert it too, but :)