throughbirds: i think a very basic implementation of the dashboard that given a room, retrieves the latest messages, stores it in the database and displays it to the user would be good. julian45 might have more thoughts.
2025-04-01 09106, 2025
vardhan_ has quit
2025-04-01 09122, 2025
SigHunter has quit
2025-04-01 09129, 2025
SigHunter joined the channel
2025-04-01 09120, 2025
mayhem[m]
lucifer: kiss ran out of disk.
2025-04-01 09135, 2025
mayhem[m]
a zstd compress is running, which will likely fail now.
2025-04-01 09139, 2025
mayhem[m]
I'll try and remove an old dump
2025-04-01 09122, 2025
mayhem[m]
site back up now.
2025-04-01 09145, 2025
lucifer[m]
[@mayhem:chatbrainz.org](https://matrix.to/#/@mayhem:chatbrainz.org) I see. I'll look at it in a bit.
2025-04-01 09152, 2025
derat[m]
reosarevok bitmap aerozol seeking your opinions on how release editor documentation bubble text should be phrased for https://tickets.metabrainz.org/browse/MBS-5944! some options:... (full message at <https://matrix.chatbrainz.org/_matrix/media/v3/download/chatbrainz.org/ugjZIGIvFIUmGZNbVViYxtNL>)
2025-04-01 09153, 2025
BrainzBot
MBS-5944: Release editor: add documentation bubbles for extra fields
2025-04-01 09131, 2025
derat[m]
(the second and third ones are existing strings. i think the second one needs a hyphen in "identically-named" and the third one shouldn't have a comma splice)
2025-04-01 09111, 2025
reosarevok[m]
Well, the third probably could be two sentences instead. I don't have a clear preference but it would make sense to make them somewhat consistent. I think we might have yet another case where it's just "Blah blah blah blah ({doc_link|documentation})" as well?
2025-04-01 09152, 2025
reosarevok[m]
I'm not sure of the second needs a hyphen or it's more of your love affair with hyphens, but :D
2025-04-01 09119, 2025
reosarevok[m]
Ideally we'd come up with one style we like and try to adapt all to it - I'd love to see what aerozol thinks would be the best option
2025-04-01 09113, 2025
derat[m]
oh yeah, i guess there's also this one:
2025-04-01 09113, 2025
derat[m]
4. Review the **list of packaging types** for help.
2025-04-01 09114, 2025
reosarevok[m]
💖✨️consistency✨️💖
2025-04-01 09152, 2025
derat[m]
/artist/create uses "Select any type from the list to see its description. If the work doesn’t seem to match any type, just leave this blank." and then shows per-option text. i don't love that approach since it increases the number of places that need to be updated if/when the guidelines change
2025-04-01 09113, 2025
derat[m]
bonus points for the "work" typo when selecting an artist type
2025-04-01 09130, 2025
reosarevok[m]
The per-option text there is loaded from the database, in any case
2025-04-01 09140, 2025
derat[m]
ah, interesting
2025-04-01 09155, 2025
reosarevok[m]
All types are supposed to have definitions, although not all of them do rn
2025-04-01 09122, 2025
reosarevok[m]
And that's supposed to be the most official one - ideally the docs would also use the same strings automatically
2025-04-01 09133, 2025
reosarevok[m]
That said, that only works for a few things, like relationships and types
2025-04-01 09145, 2025
reosarevok[m]
Since most guidelines and the like are just text on a doc page :)
2025-04-01 09111, 2025
derat[m]
there's a separate issue of displaying a bubble for the artist field. i think that it uses KnockoutArtistCreditEditor, and i have no idea how that interacts with the controlsBubble knockout stuff that's used for the regular inputs in root/release/edit/information.tt
2025-04-01 09106, 2025
derat[m]
(fwiw, apologies for all of my changes adding more knockout junk that'll make it harder to move the release editor to react :-/)
2025-04-01 09158, 2025
reosarevok[m]
That bit is where bitmap can certainly help better :)
mayhem: no extra files or anything this time, we need more space to do the full dumps. the uncompressed data is too large even if after zstd compression it takes only 200250 GB.
I think we may need to upgrade servers then. let me check options.
2025-04-01 09130, 2025
mayhem[m]
another alternative is to add an extra drive to the existing server. it wouldnt need RAID-1 for building dumps, right?
2025-04-01 09148, 2025
lucifer[m]
mayhem: don't think so.
2025-04-01 09158, 2025
mayhem[m]
ok.
2025-04-01 09126, 2025
mayhem[m]
zas: do you know how we can find out what options we have for adding a drive to kiss? we need another TB of non RAID storage.
2025-04-01 09116, 2025
zas[m]
let me check
2025-04-01 09150, 2025
zas[m]
Not sure, this type of server isn't listed anymore, so we have to contact Sales to know what options we have. But currently it uses 2x1.92TB nvme drives. I guess we can add more, but I fear the price will be high (as most options at Hetzner). What are we targeting to size-wise?
2025-04-01 09100, 2025
zas[m]
We can convert RAID1 to RAID0, if we can afford the risk.
2025-04-01 09127, 2025
mayhem[m]
1TB
2025-04-01 09152, 2025
mayhem[m]
and we can afford the risk, since this is for making data dumps. if the drive fails, we run a new dump, no big deal.
2025-04-01 09114, 2025
mayhem[m]
and another server looks to be more expensive and not needed, so just a single disk for now would be good.
2025-04-01 09117, 2025
zas[m]
Is a complete system erase possible for this machine? (I can try to do it while keeping data, but ... it may fail)
2025-04-01 09129, 2025
mayhem[m]
even for just adding one drive?? we want to leave the existing drives alone.
2025-04-01 09110, 2025
mayhem[m]
I think if we need to prepare for a whole system erase, then we might as well just upgrade. its easier.
2025-04-01 09104, 2025
zas[m]
ah, I thought you wanted to reclaim 1Tb from existing drives using RAID0. My bad. We can ask for one more drive.
2025-04-01 09113, 2025
mayhem[m]
great
2025-04-01 09136, 2025
zas[m]
I sent a request to Hetzner to see which options we have on this type of server.
2025-04-01 09109, 2025
tykling
they might even be able to do you a usb stick, the biggest in the shop are 64gb but they do have larger ones
2025-04-01 09147, 2025
mayhem[m]
and we'll be able to create one data dump every 6 months with a USB stick. :)
2025-04-01 09157, 2025
tykling
:)
2025-04-01 09159, 2025
bitmap[m]
<derat[m]> "reosarevok bitmap aerozol..." <- I'm not sure strings like "this is the release name" or "please enter the barcode" are very useful, they're just repeating the field name back to the user. so I'd prefer something like the disambiguation one where it's a short summary of what the field is for + a doc link, or alternatively just a bare doc link
2025-04-01 09149, 2025
bitmap[m]
agreed about the hyphen + comma splice though I don't have a strong opinion about the former :)
2025-04-01 09152, 2025
derat[m]
bitmap: thanks! if aerozol agrees, i'll go with that style for all of these
2025-04-01 09105, 2025
derat[m]
<derat[m]> "there's a separate issue of..." <- bitmap: do you have any ideas about this? it seems pretty gnarly at first glance, so i might just leave the artist autocomplete bubble-less for now
2025-04-01 09108, 2025
pite joined the channel
2025-04-01 09147, 2025
vardhan_ joined the channel
2025-04-01 09152, 2025
bitmap[m]
<derat[m]> "bitmap: do you have any ideas..." <- hmm, the simplest option is probably just to use [jquery event delegation](https://learn.jquery.com/events/event… (maybe install a handler somewhere in release-editor/init.js) and call `bubble.show()` in there if the AC input is focused
2025-04-01 09143, 2025
bitmap[m]
although you can't focus that field if a complex AC is entered 🤔
2025-04-01 09108, 2025
derat[m]
thanks, i'll look into doing that! my main goal with this is just to provide guidance to beginners, so i'm happy enough as long as the bubble gets shown when an initially-empty field is focused
2025-04-01 09111, 2025
lucifer[m]
mayhem: btw on the similarity data front, i replaced recording_mbids with recording_id's that reduced memory usage 4x and storage space usage 8x.
2025-04-01 09109, 2025
mayhem[m]
smrt. I do the same in faster_fuzzy.
2025-04-01 09129, 2025
lucifer[m]
i still need to handle redirects so it might give away some gains but yes lets see.
2025-04-01 09130, 2025
lucifer[m]
haha cool.
2025-04-01 09109, 2025
lucifer[m]
i have noticed that in a lot of cases parquet doesn't compresses better than zstd compressed csv's and i think it does have to do a lot with uuids vs numeric ids.
2025-04-01 09119, 2025
lucifer[m]
because 80% of our data is just that.
2025-04-01 09145, 2025
lucifer[m]
another thing i noticed is that zstd compresses ~2x better for mlhd if we order by a column.
2025-04-01 09149, 2025
mayhem[m]
and those are text UUIDs or binary UUIDs?
2025-04-01 09110, 2025
lucifer[m]
parquet treats them as text i think.
2025-04-01 09146, 2025
BrainzGit
[musicbrainz-server] 14mwiencek merged pull request #3498 (03master…mbs-12170): MBS-12170: Output containing country code for all artists in web service https://github.com/metabrainz/musicbrainz-server/…
2025-04-01 09155, 2025
lucifer[m]
or arrow at least. i need to look it up again but last time i checked due to reason it was being treated as text.
2025-04-01 09128, 2025
mayhem[m]
yeah, I think PG is one of the few tools that does treat them as binary
2025-04-01 09153, 2025
lucifer[m]
yeah spark doesn't have a uuid type either.
2025-04-01 09157, 2025
lucifer[m]
i am curious how better the stats/dumps processing for spark can be if we switch to ids instead of uuids and then before sending the stats back to LB, switch id to mbid.
2025-04-01 09130, 2025
lucifer[m]
the unmapped listens create troubles though so...
2025-04-01 09136, 2025
mayhem[m]
I would expect it to go as well as the instances you've already seen
2025-04-01 09115, 2025
mayhem[m]
maybe consider a dense encoding of UUID?
2025-04-01 09103, 2025
lucifer[m]
hmm could be useful but those aggregating queries also have recording_name etc in group by because mbids are optional for stats anyway.
2025-04-01 09132, 2025
m1gr has quit
2025-04-01 09158, 2025
m1gr joined the channel
2025-04-01 09123, 2025
minimal joined the channel
2025-04-01 09146, 2025
mayhem[m]
lucifer: signging up for a deezer account was a bad idea. much spam and the unsubscribe page is 404.
2025-04-01 09103, 2025
lucifer[m]
Rip. I don't like deezer too because I can't even view their developer docs without vpn and even that is aggressively blocked.
2025-04-01 09155, 2025
suvid[m] joined the channel
2025-04-01 09156, 2025
suvid[m]
<lucifer[m]> "Rip. I don't like deezer too..." <- I think that's cuz deezer is not present in India right?
2025-04-01 09132, 2025
lucifer[m]
Yes.
2025-04-01 09100, 2025
vardhan_ has quit
2025-04-01 09129, 2025
Sophist-UK joined the channel
2025-04-01 09106, 2025
aerozol[m]
derat: bitmap reosarevok Agreed with bitmap that I would cut out the parts that are just repeating what the field is. Unless it addresses a common problem - for instance if users often get confused about digital barcodes: “Please enter the barcode from the physical packaging or the digital storefront of the release” (just an example, I imagine reo will have an idea of what fields new users really struggle with/what could benefit
2025-04-01 09106, 2025
aerozol[m]
from being spelled out)
2025-04-01 09123, 2025
aerozol[m]
For tutorial/guide stuff I would always use short sentences if possible. It reads less natural, but for those with bad English it is much easier to read. You can google “plain English” for the philosophy behind that. I have also embedded it into our writing styleguide, which everyone has forgotten about I’m sure :P https://github.com/metabrainz/design-system/blob/…
2025-04-01 09111, 2025
aerozol[m]
All the “pleases” are definitely consistent with reosarevok polite style throughout our guides and docs! 🥰
2025-04-01 09135, 2025
DarthGandalf is now known as [LLM]DarthGandal
2025-04-01 09152, 2025
mthax_ joined the channel
2025-04-01 09155, 2025
mthax has quit
2025-04-01 09140, 2025
lucifer[m]
zas: hi! let me know when you are around tomorrow, i would like a CCX63 VM with a 5 TB storage volume.
2025-04-01 09155, 2025
SigHunter has quit
2025-04-01 09131, 2025
SigHunter joined the channel
2025-04-01 09139, 2025
mayhem[m]
<lucifer[m]> "zas: hi! let me know when you..." <- Back in my uni days we toured amdahl, a fault tolerant mainframe computer manufacturer. they had a silo with a robot arm and gazillions of tapes and tape readers... it was quite impressive. and it stored 1/4TB. and you just casually asked for a 5TB machine, like nothing. lolololol
2025-04-01 09135, 2025
lucifer[m]
😆
2025-04-01 09124, 2025
Kladky has quit
2025-04-01 09113, 2025
nelgin
Is it easy/difficult/imposssible to get the actual cover art into the docker image?
2025-04-01 09117, 2025
kepstinbrainz joined the channel
2025-04-01 09117, 2025
kepstinbrainz
nelgin (IRC): metabrainz doesn't store the actual cover art, that's all hosted by archive.org - my understanding is that it's far too much data to reasonably put into a docker container.
