#metabrainz

monkey: i think mayhem will need to check spotify dashboard and update the uris there if needed. iirc test.lb is using a separate spotify app from beta and prod.

one sec

Yep, was just writing that I don't have access to the spotify account

:)

nice!

not a million, but not 10 either

looks like we never updated that after adding the new connect page

Yep. mayhem the callback URI should be: `https://test.listenbrainz.org/profile/music-services/spotify/callback/`

this is the one for prod:

Please and that you :)

ah, interesting. I guess we're using the same app on both prod and dev?

thank you*

no, but we could.

updated, monkey

Thank you

maybe remove the test callback urls for the prod instance? to reduce possible future confusion

monkey: you'll have to write spotify gdpr support that you want timestamps of the export in milliseconds. if they still send in seconds then write back to them emphasizing that its seconds but you want in milliseconds.

works now.

alastairp: agreed.

done

OK, will try, thanks lucifer.

great

alastairp: yeah, it had taken some back and forth but i was able to get ms precision timestamps last time.

lucifer: oh! wow!

and you're not even from europe :)

Alternatively, I might just sort the playback history and try to import all the ones that are before my first LB listen

this is the exact email i had sent.

alastairp: spotify needn't know that :)

Thanks

oh i sent this first email to privacy@spotify.com not support. rest of comms were with support.

alastairp: using linux, yeah

lucifer: monkey: removing listenbrainz/webserver/static/dist worked

took me quite a while to figure out that Picard Windows code signing was failing because MBs US address has changed :( Assumed it is me too stupid to use openssl properly

hurray for cryptical error codes instead of human readable messages

ohhh, that sucks.

and on top of that, our address has to change again, since our mail forwarding company moved. 🤦

ha :) but the certificate is done now and valid for a while. the problem here was just that for the windows app package you need to set the publisher, and that must match the certificates subject. but publisher was hard coded

monkey: played 8 tracks without stopping!

Gosh, the bar really is low !

And I am definitely seeing working token refreshing, so I think that's that particular problem solved ! W00t!

alastairp: https://api.listenbrainz.org/metadata/recording... this 410 Gone redirect so that people can't use the API calls on lb.org (instead must use api.lb.org) -- is that managed from the gateway, do you remember?

yes, it's on the gateway

zas: ^^ if you're around.

mayhem: it should be https://api.listenbrainz.org/1/metadata/recordi...

the /1

ah

lucifer: yes. I just wanted to remember where and how we did the 410 Gone tricl.

ah ok

trick. since art.lb.org needs the same trick done.

yes, its at gateway level. https://github.com/metabrainz/openresty-gateway...

I just found that too :)

thanks. :)

Looks like it's not just me being a little slow with the spotify SDK token refresh thing. Their own example application gives bad advice as to how to set the player up: https://github.com/spotify/spotify-web-playback...

damnit. player stopped. network tab not open.

lucifer: not for LB#2226, but a later one.... I think for the similarity stuff we should exclude the seed artist_credit from the similar recordings

Add similar recordings lookup: https://github.com/metabrainz/listenbrainz-serv...

because we're getting a lot of hits for the same artist that I just filter out -- invariably they are the other tracks from the source album.

monkey: do you want the HAR file if the player fails to move on the next tracks after I pressed the Next button?

Yes, with the indication of both tracks please (the one you were listening to and the 'next' track)

from Acid fro Nothing (I wish!) to .Blank

never mind me. I'm a total dumbass.

not on test.lb.org :facepalm:

Hah! I almost asked but thought better of it.

mayhem: pong, what is it needed exactly?

this commit implements an exclusion so that people who try to make an API request off listenbrainz.org (and not api.listenbrainz.org) are given the error message and the 410 error.

we want to do the same for the new cover art page: art.listenbrainz.org

ok, let me see

so any user that tries to use listenbrainz.org/art should get a 410 telling them to go to art.listenbrainz.org

and listenbrainz.org/1/art should need to go to art.listenbrainz.org/1/art

thx

actually the patch you linked tell people using api.listenbrainz with /<number>/ to use listenbrainz.org

this is what I mean:

but there's another one

k

we should probably setup the 2 equivalents for the art case.

art.listenbrainz.org isn't configured on gateways

it points directly to wolf atm

correct. once we are ready to deploy this PR we will move it to gateways.

it would be easier to do the reverse, because I need names (like prod.listenbrainz-web), the rest is basically a copy'n'paste. Do you know which target names it will use on prod?

kiss, I would expect.

yes, but which container?

should we do this after we deploy this PR and move the DNS entry?

one sec.

it will be easier for sure, as is it would be hard to test

ok, I'll come back to you when we deploy this.

(since files will not be used until target containers are up)

but it will be listenbrainz-art-{test|beta|prod}

the rest is basically a copy'n'paste of what we have for LB

ok, cool.

thanks!

I'd recommend to stick to (prod|beta|test).service

maybe prod.listenbrainz-art ?

yes, since what the gateway uses is slightly different than the actual container names.

yvanzo: can you check sir?

mayhem: i guess we can add a flag for that filtering? or should we do it always

zas: on it

zas: all good

yvanzo: thanks for the logs, i'll look into it. btw is this a blocker for doing a new releaes?

lucifer: yes

ah ok 👍

yvanzo: from a quick search of the db error message, it seems benign. and debug logs of sir do not seem to indicate any issues either? did it crash or completed normally?

for instance, https://www.reddit.com/r/PostgreSQL/comments/43...

it doesn't complete at all

i see

I mean it doesn’t return but events seem to be all processed.

hmm, weird indeed. and it only happens for events?

no, I mentioned it can happen for some others.

events, release groups, urls...

oh ok.

it seems that Postgres log lines more or less match loading relationship in SIR debug.

the PG log lines seem unrelated to me. my intuition is that some thing related to multi processing is hanging up but could be wrong. i'll try to reproduce the issue and see if i can fix it.

..

lucifer: do it always. I can't see how that data is any good.

i see. we could also try doing it on spark level then?

spark level would be best.

cool, will do

there is no need to move those lines around, they are useless.

and can you please run the following 4 days sets for me? d 1095, s 100, t 5; d 730, s 100, t 5; d 1095, s 200, t 5; d 730, s 200, t 5;

sure

th

x

time to see what the session does. I keep getting 2-3 tracks of heavy metal in this test playlists. when it doesn't belong.

zas: It seems that SIR is 10x slower than usually but I cannot explain why atm.

(Restarting it didn’t help either.)

just 10x?

atj: Actually more 8x or 9x but I round it up.

when did this start, do you know?

roughly

atj: 14:50

+-2min

mayhem: heavy metal always belong

Oh, btw. Meeting is in ~7 minutes for anyone who doesn’t follow EU DST. bitmap, akshaaatt, lucifer, CatQuest, Pratha-Fish, Shubh, yellowhatpro, ansh, riksucks, … uh… and anyone else I may have missed.

atj: GMT/UTC