#metabrainz

reosarevok: MBS-4073 seems fine to me for next schema change

MBS-4073: Highlight pending changes to labels in sidebar https://tickets.metabrainz.org/browse/MBS-4073

+1 to closing MBS-4107

MBS-4107: I18n - Internationalise the MusicBrainz website https://tickets.metabrainz.org/browse/MBS-4107

the Number.prototype.toLocaleString stuff should resolve MBS-4118, though I'm certain we don't use it everywhere we should and it needs to be tested in more languages to be sure

MBS-4118: Localise number formats https://tickets.metabrainz.org/browse/MBS-4118

the vote reports thing seems like a nice idea. why don't you think unreviewed edits should be marked as such more visibly in the listings? (just curious)

re: speeding up the applied edit count predicate, an index on (editor, status) might help there but would need to test, perhaps on queen

spellew: thanks! I'll take a look :)

mo’’in’

reosarevok: on MBS-4219 -- not yet, but maybe early next year?

MBS-4219: Show similar collections on collection pages https://tickets.metabrainz.org/browse/MBS-4219

bitmap: I don't necessarily think it should not be marked, but on the other hand, given we still don't even show the vote tally in the listings (just "My vote: None"), that's quite a jump. Maybe showing the tally as 0 Yes 0 No would be enough?

I guess if we do that we can also add like a special border or something to "0 Yes 0 No" :)

ruaok, bitmap, yvanzo, iliekcomputers: pink is ready (meant to replace queen), I still have few minor things to fix, but the system is ready, docker is installed, monitoring is working (I still have to figure out how to collect data from adaptec hardware raid). It required many small changes since hardware isn't exactly the same (mainly network interfaces and hardware raid).

please test if you can log in

bitmap: you can start configuring it to replace queen (that should be removed before we can order bowie's replacement)

when floyd (new bowie) will be plugged in: I'll set it up, then we switch to pink as main database server, move stuff from bowie to floyd, when done, we toggle floyd becomes master again

confirmed. I logged in and can sudo.

:)

but can you log in and do judo? 🤔

"(meant to replace queen),"

whaaaaaat? nooooo

ಠ_ಠ it's going ot b one of those days isn't it -__-

CatQuest: queen and bowie are being replaced due to needed hardware upgrades. Being replaced by pink and floyd respectively.

Freso: I upgraded discourse

alastairp: ping

I love New servers

Mostly because of new server names

pink.metabrainz <3

ruaok: hi

how long do you need?

maybe 10-15 minutes?

perfect, let's do it

I think I may have sorted my question now, but a headcheck would still be good.

ok.

as you might recall, I run a query on the MB data to fetch tracks that have release type album and insert that into the MSB DB.

then I parallel walk both that table and the MSB recordings in sorted order to find matches.

that part is working nicely and artist clusters are being voted on. all good.

my concern now is to pick matches against recordings so that all MSB recordings with the same text end up on the same MB release.

zas: Offending key for IP in .ssh/known_hosts (line matches bootsie)

yvanzo: delete the old one. they re-used the IP.

alastairp: because I want to avoid scattering MSB recordings across disparate releases for a given artist.

so, my current thinking is to find the first released release for each release group and match against that if the MSB recording does not specify a releae.

from MusicBrainz you get only a list of recordings, no release names?

the thinking is that people who want to listen to other people's listens would most likely want to listen to the album release recording as the first release.

I'm working now to get release names, in part to ensure that MSB recordings get clustered on one release and not across a pile of them.

my previous approach did not involve release names at all.

and that approach matched recordings in a willy nilly manner that doesn't make for nice data clusters.

(or a nice user experience)

> 12:31 PM <ruaok> my concern now is to pick matches against recordings so that all MSB recordings with the same text end up on the same MB release.

is this messybrainz recordings without a release name provided, or those ones too?

without.

next I will need to consider ones that have releases provided.

right, sure

but I find that taking small steps is making the whole thing manageable in my head. otherwise it quickly gets out of head and out of hand.

does that approach make sense?

do you want to use context to help here? e.g. user a listens to track 1, 2, 3 in order, find where they appear in a release?

I suspect that's a pretty complex process though

oh, drat. I guess messybrainz doesn't have temporal info :(

but yes, getting the first release that a track appears on is a good safe start

it's what I would do

ruaok: thanks

zas: confirmed too (login and sudo)

alastairp: yeah, MSB is a very strange slice of data to work with.

we did a great piece of handwaving back then in london. :)

ok, then I will work towards that and see what I get out of it.

"let's do something and hope that it's useful for us"

I suspect that this first pass is going to match a lot of data. once I am happy with that, I can look at other data and find candidates for merging into the clusters that have been found already.

THAT.

it is useful, but tricky.

not quite https://musicbrainz.org/artist/5bf64d94-efd9-43... , but still.

ruaok: are you just matching track names, or combining artist too?

both.

great

one single value to too shit.

Are you matching tracks that don’t appear on type:albums at all? (Or one step at a time and this will be done later?)

later.

:C

phw/picard#119 (master - ddb628c : Sambhav Kothari): The build passed.

Change view : https://github.com/phw/picard/compare/160c94417...

Build details : https://travis-ci.org/phw/picard/builds/559387352

oh alastairp....

you do not need access to the database on frank for running aidanlw17's stuff, right?

no, I was planning on making a duplicate database

great.

so that we can play around with it before touching prod

4TB disk ok?

32GB of ram or 64gb?

likely, how large is the database now?

how much ram does frank have?

you're asking me??

iliekcomputers: got an estimate of the size of the ab database?

frank has 2TB drives.

so 4 is plenty

and how much ram does frank have?

128GB of ram.

I trust annoy is ram intensive, no?

I almost spat out my tea

really?

nah, annoy uses mmap

so it's all on disk

131719944 according to free

unless linux caches it

here is where I am going with this.

if it's not a huge increase in price to go to 64, then that'd be ok

hetzner has server auctions.

no minimum duration, no setup fees.

(what's the catch?)

the catch is that you can't do many add ons or order multiples of the machines.

which we don't care about.

oh wow, sounds good

and for 2 months -> 100›€? venga!

zas: take your time with the server migrations. I'll get alastairp setup with server bidding outside our rack.

ah right, because we don't need frank db means that it doesn't have to be in the rack?

yerp.

great, thanks

I didn't think of this before.

can you do the bare machine setup on this beast?

sure