#metabrainz

Can I get some feedback on using one of these along with LB update posts monkey? I will add our ‘bubbles’ with version numbers and so on

p.s. I’ve started treating these as a set, which is very satisfying!

They look great! Good job 👍

Thanks Maxr1998!

The heavy lifting is done by mockup templates (all with commercial use licenses ofc), but still a lot of hours to make it come together

bitmap: yes that would be great

I don’t know if this is better mayhem, what do you think

Looks great when the flexbox gets wider though

[data-set-hoster] 14dependabot[bot] opened pull request #17 (03master…dependabot/pip/pydantic-1.10.13): Bump pydantic from 1.10.8 to 1.10.13 https://github.com/metabrainz/data-set-hoster/p...

[listenbrainz-server] 14dependabot[bot] opened pull request #2849 (03master…dependabot/pip/pydantic-1.10.13): Bump pydantic from 1.8.2 to 1.10.13 https://github.com/metabrainz/listenbrainz-serv...

yvanzo: started looking into https://github.com/metabrainz/musicbrainz-docke...

building images for now

aerozol: those looks great! Slight preference for version A but they all work. Nice looks!

bitmap, reosarevok: still on my way back home at the moment

Ok, no worries :)

moooin!

lucifer: where are these reports?

aerozol: those looks great! <3

bitmap, reosarevok: Looking at the recordings with the longest names, it seems that 1024 characters can be a more than reasonable hard limit. it wouldn’t require any change to the database schema either.

Most of these are intentionally crafted by database trolls, and I’m totally fine with not wasting MB resources at supporting these in the `name` field. Their full “title” can be added in annotation if editors think that it is worth their time.

Hmm

How come it doesn't require schema changes? Wouldn't we need to alter the column to have a lower char limit?

Or would you just limit it on the Perl side?

Some others are obvious mistakes such as the tracks of https://musicbrainz.org/release/703d01da-5a67-4... which have several track titles for each track.

When I first looked at the related MBS-13536, the longest legitimate correct title had less than 500 characters and was a classical track, that was probably listing all the works instead of just the title.

MBS-13536: Too large title causes Postgres index error https://tickets.metabrainz.org/browse/MBS-13536

reosarevok: Just through edit validation in Perl indeed. I agree that having an additional constraint in PG can be safer too, but it isn’t strictly at first, and it is much simpler than modifying indexes anyway.

O

*Ok :)

I think that sound effect thing might actually be correct, but it's not super important in any case

It isn’t, see https://www.discogs.com/release/12423143-Sound-...

Ok :)

It might also be a Discogs’ editor choice, who knows, but the track titles in MB make no sense, it looks like a bad copy-paste.

Yeah, I was wondering because these sound effect albums sometimes have very weird track titles :)

But you're probably right, it looks extra weird

We can also have some guidelines saying that too long titles should be considered as liner notes and be put in annotations.

yvanzo: I just restarted sir-prod, it was errorring

1024 characters is enough for anyone

Even for a few tens of potentially legitimate recordings in 30M+ of recordings, it isn’t worth requiring all the tools accessing the MB DB to support more than 1024 chars.

zas: Thanks!

yvanzo: FYI we're probably going to end up running the new Solr cluster on arm64 servers

atj: Amazing!

they're about 30% cheaper and we get 2x the cores

I won’t be around most of the day but I will spend the time I have at patching edit validation.

i've got a mixed 7 node cluster handling 450req/s without too much problem

that was with 6 x86 and 1 arm64 node

yvanzo: you, zas and i should probably have a chat next week as I think we're approaching a production ready setup

yup

we still need to round some corners, but we're getting there

yvanzo: I think that sounds sensible really

yvanzo: huh, not sure what went wrong (but I did not have a fully cleared setup, so maybe that's part of it):

(while running sudo docker-compose run --rm musicbrainz createdb.sh -fetch)

atj, zas: Yes, the plan is to reset solrcloud data from the production database, to set up a 2nd sir instance to live feed it, and to make beta.mb.o use it for search.

yvanzo: i disabled the documentcache entirely for testing because of the lack of updates from test.mb.o

so it wasn't ever getting flushed

to be honest i'm not sure it's even worth it with a 10 second soft commit time

reosarevok: is there a proposed date for the next mbs release?

Well, we don't have a lot to release rn I think, ideally we'd put the eaa in beta for a little bit, let people test it, then put that out

But if there's something you want out in prod, we could do a smaller release too, that's fine

reosarevok: is that test setup? this error can occur when `carton exec` is run but `cpanfile` is not found. you can check that this file exists and that `carton` is run from the right directory. Or maybe some environment variable should be set (you can try to pick it from mb-server repo)

I just dropped the containers, checked out your branch and started following the readme :)

The current beta requires perl 5.38, it can’t be released in prod before the schema change.

For patching edit validation, we have to target the production branch directly.

atj: ok, in that case the proposed date is May 13 :)

reosarevok: so May 13th for schema change release?

Yeah

which includes the upgrade to Solr 9?

bitmap: The MB mirror on wolf is up-to-date, did you fix replication packets already or is it just not affecting MB Docker mirrors?

atj: yes

3 weeks then 😬

bitmap: Or is it just affecting the latest dumps? (let me dream)

I tried the lb playlist generator with country:(india) (or japan or anything) and all I get is

"LB Radio generation failed: cannot parse prompt: 'Expected {{{'tag' | 't'} Suppress:(':') ... blah blah hugess string"

idk. i tried with and witotuh coutes, i tried replacing with tag isntead idk

it doesnt seem towork?

it'd be good if it actually had copy-paste seartch urls too!

KassOtsimine (IRC): The country stuff is new and I think it was only deployed to beta for a little bit, but has been replaced since

hmm alright, but tag:;(something) brings the same error

-;

Brings a different error for me, but a known one

monkey[m]: I looked into migrating the class components to functional components. You can only use wrapper with PlayerPage, not the other components as they have the listens saved in states which are updated based on user interactions. I'll push that change directly in the brainzplayer-spa branch.

for simply tag:(trip hop)

IK get

LB Radio generation failed: Cannot fetch recordings for tags. {"code":400,"error":"pop_begin param is missing"}

ansh (IRC): OK, that's one less component to deal with :)

🤷‍♂️

yep KassOtsimine (IRC) that's ruight. Known issue, I reported it yesterday too

oh! if it's known

Well, I guess it's in a known broken state

The artist prompts work

just clarifying: I'm ona modern browser now

Yep, this time it ain't you ;D

Enjoying the modern web?

Woww KassOtsimine! Hopefully it's not too annoying!

what are you all going on about? thisis theothe computer. i already toldyou the new computer isn't owkring :/