#metabrainz

bitmap: ruaok: bono was slow probably because its running a dev instance. in dev, the js files are almost ~10 MB each. in prod, those are shrunk down to ~800 kb each + plus get cached.

right, I figured they just weren't minifed

MB's are probably >10 MB in dev if you include source maps, though we output those to a separate file

makes sense

btw MB is using ES5 or ES6?

we compile everything to es5

in the actual source code we use stuff from es2020 afaik

ah nice

Hi lucifer!

hi!

So to be able to run the color browse branch locally, I would have to use port forwarding to bono.meb?

Or how does the steps go?

do^

you can bring up an instance on bono.meb. and then access it using the public url.

Okay understood. How do I bring up the instance?

./develop.sh up

But that starts it in localhost, right?

How do I bring it up in bono?

you'll probably get an error currently, because my containers are up. i'll take those down.

run it on bono. i mean ssh bono.metabrainz.org, git clone lb server, checkout branch, run ./develop.sh up.

Interesting so I wasn't cloning the project in bono to do stuff. Got it

What I was thinking was that I could mirror my local work to bono somehow and access the stuff there. But I will have to setup everything in bono again right

Great, I see it's already allocated.

You must be using it I suppose

it might be possible to mirror your setup but don't think it will be easier than manually copying.

yes i'll bring mine down

Thanks

done

"ERROR: for web Cannot start service web: driver failed programming external connectivity on endpoint listenbrainz_web_1 (dc64de3e4658e87683554047ac592dec4a1ba603a6a86a231d35485d0c6e5b1a): Bind for 0.0.0.0:80 failed: port is already allocated"

Still do see this

which branch?

color browse only

let me check, it had the workaround until yesterday

Oii I'm up

yeah its still there. can you recheck the branch?

Yes working

yup i see the containers. nice

Got it now. Need to do some more stuff.

Let me know if you're working rn lucifer

If yes you can have the space and continue your work

you may face errors on bono often because of outdated config.py. so do remember to keep it up to date when using bono.

sure, if you're not using it, bring those down.

Right. And how do I access the clone in bono on my ide?

depends on the ide but it'll probably ssh forwarding

Cool

Thanks!

Please ping me once its running on bono :)

lucifer I found a gem finally

we can use vscode to edit files on the server directly

Damn convenient

Right, and that comes with port forwarding :)

lucifer: is https://community.metabrainz.org/t/musicbrainz-... a known issue?

don't think so. i'll check.

I was unable to login to mb on bono with my containers up

lucifer is there an additional setup required for the authentication?

akshat, not sure what you mean

what do you mean by login to mb?

Test bono.meb

Try sign in

http://bono.metabrainz.org:8081/ you mean this error ?

Yes

config.py needs to be amended probably

outsidecontext, rdswift: I was told that https://picard-docs.musicbrainz.org/en/variable... is a bit outdated and the relevant checkbox is now "Enable track relationships". If so, can we update it? :)

outsidecontext, zas: in a related question, does Picard have any way currently to get relationship data for releases that go over the 500 recording limit? We really should figure out a way of doing that - if it needs a new endpoint we could talk about it with bitmap and yvanzo.

(without that, the payoff for people actually improving the data on huge releases is super low :) )

afaik Picard doesn't do anything special, and depends on the web service to provide data, if it is limited server-side, then Picard can't rely go over that. Now I'm not sure what's this 500 recordings limit. I wish we had ws "paging", and no limit at all.

Yes, that's limited server side. The only way around would be currently to load the data piece by piece afterwards, probably per recording. That was proposed before, but it obviously would be super slow (8+ minutes with the 1 request per second limit)

So that's mostly a server side issue. Having the ability to get the recording data for a release paginated would be a way around that I guess, e.g. if the endpoint only returns 500 recordings per call. Even for most huge release it would still be only 2 or 3 calls I guess

Yeah, an option I could see is having a special endpoint that returns just relationship data for the recordings/works, given a release MBID

I guess that's kinda what you were suggesting :)

I wanted to make sure Picard didn't currently implement a "check every recording" option or anything

But yeah, I guess that'd be super annoyingly slow

bitmap: I'd really like some feedback on this when you're around ^ :)

reosarevok: not sure about that, if the endpoint loading release data with many recordings + works and their relationships is slow a separate endpoint for loading the recordings + works and their relationships will likely just suffer from the same issue. As I understand the problem with loading the release is really the data on recording and work level

I meant, having that paginated

So you get the recordings without rels in the normal request (with *some* notification that rels were not returned)

Yes, makes sense

And then you can get a paginated endpoint that is just recordingmbid: rels data somehow

https://tickets.metabrainz.org/browse/PICARD-657 has some discussions on options on Picard side, but the core issue is I think https://tickets.metabrainz.org/browse/MBS-3680

PICARD-657: Gateway Timeout with large releases

MBS-3680: Error 502 Bad Gateway loading certain releases from the webservice

moin!

the bot got 55k images indexed before getting a 403 axe.

about handling of huge releases, I guess bitmap has few ideas. This is something we could discuss during the summit. It seems to me that will be more and more a problem since we get more and more metadata.

zas, bitmap : thanks for the CAA setup. working great. now running with 16 threads.

patton at less than 20% cpu.

ax51 are real workhorses

I love the AMD cpus. I have one on my desk at works and its lovely.

Moin!

moin!

I'm just gonna leave this here in case it's usefum: https://www.npmjs.com/package/react-colour-wheel

useful, even

monkey: one of the keys is that we want to specify our own colors. an evenly spaced color wheel doesn't work well.

because the cover art color space is anything but evenly distributed.

You can specify your own colors with this library

oh. very good.

akshat: lucifer ^^ have you considered this?

I thought the original design was also evenly spaced, no?

monkey: the colorspace, not the colorwheel.

Ah, I see :p

as you might imagine all of the metal images are black.

so, you need to have several color slices to pick out different shades of doom metal

*surprised face *

exactly.

there are other bunched up spaces as well.

The nice thing with this library is it allows you to refine the selection (see the demo gif on the page)

moiin monkey

ohh, that could be very nice.

zas: patton is load 5-6 now, but holding steady.

yes, I directed quite a lot mb ws traffic to it

but I think the extra load is ok

ok, cool. I saw the load hovering 3 before I got started.

yeah, looks ok.

monkey: one thing I wanted to discuss with you.... but.. are you the in office today, perchance? this will be easier in person.

So say you select black, you'll then have a choice of 10 dark gray/black colors

that colorwhell really does look very useful.

and it indicates the color you picked. could be quite good.

perhaps not as sexy as the one we had, but this is more functional