#metabrainz

From inside the container https://www.irccloud.com/pastebin/Bgc4SxNP/

root@rex ~ # redis-cli -h 10.2.2.60 -p 13062 hangs

equivalent command on kiki gives a prompt immediatly

so it works, but not from the machine itself

what is the docker command you used to launch the container?

it uses https://github.com/metabrainz/docker-server-con...

the version of keydb is more recent, we can try with the older version

thanks, I'm just checking something

atj: we should try with older version, it can be related to a change in 6.3.x versions

alright, give me one sec

no difference

it works from inside the container https://www.irccloud.com/pastebin/lEAMl9Te/

so that's a network-related issue

I can run nc on rex and connect to it on rudi over the internal interface, everything seems fine

does it work with redis-cli?

yep https://www.irccloud.com/pastebin/CLZblInF/

let me just try inside both containers

[musicbrainz-android] 14dependabot[bot] opened pull request #152 (03master…dependabot/gradle/androidx.test-runner-1.5.1): Bump runner from 1.5.0 to 1.5.1 https://github.com/metabrainz/musicbrainz-andro...

[musicbrainz-android] 14dependabot[bot] opened pull request #153 (03master…dependabot/gradle/androidx.navigation-navigation-compose-2.6.0-alpha04): Bump navigation-compose from 2.6.0-alpha03 to 2.6.0-alpha04 https://github.com/metabrainz/musicbrainz-andro...

[musicbrainz-android] 14dependabot[bot] opened pull request #154 (03master…dependabot/gradle/androidx.compose.material3-material3-window-size-class-1.1.0-alpha02): Bump material3-window-size-class from 1.1.0-alpha01 to 1.1.0-alpha02 https://github.com/metabrainz/musicbrainz-andro...

[musicbrainz-android] 14dependabot[bot] opened pull request #155 (03master…dependabot/gradle/compose_version-1.3.1): Bump compose_version from 1.3.0 to 1.3.1 https://github.com/metabrainz/musicbrainz-andro...

[musicbrainz-android] 14dependabot[bot] opened pull request #156 (03master…dependabot/gradle/androidx.compose.material3-material3-1.1.0-alpha02): Bump material3 from 1.1.0-alpha01 to 1.1.0-alpha02 https://github.com/metabrainz/musicbrainz-andro...

atj: can you summarize what's working (or not) and what you did?

I'm scratching my head

good, concise summary!

!m atj

You're doing good work, atj!

:D

atj: I suspect something with docker

yes, same here

there is no proxy process running, dockerd is listening on port 13062

ok, let's try to restart it on rex

"iptables: false" related?

alright, I'll restart docker

but let's see

monkey: give me two sets of 2 tracks and let my alg compare the two.

no change

level=info msg="There are old running containers, the network config will not take affect"

one set should be "these ought to be equal" and one should be "heck no, these should not be equal"

atj: yes, see above

let's reboot rex, just to rule out this (or not)

🤣

zas: ok I'll reboot rex in 10 seconds

no change

the error messages on rudi changed while it rebooted

Error condition on socket for SYNC: No route to host / Error condition on socket for SYNC: Connection refused

just going to try binding the container to a specific IP

ok

mayhem: https://musicbrainz.org/recording/b18cdfd2-b45e... and https://musicbrainz.org/recording/6844d3f9-6cb5...

Curveballs, just to see what happens

redis-cli gets stuck waiting to read data https://www.irccloud.com/pastebin/jfkliPPR/

just going to stop the container on rudi so I can tcpdump with less noise

mayhem: 5a2f56e7-bc0f-4aab-b8c9-c441633fb6b2 and 7e6790f1-886d-43c6-9e2e-bc0e491c615c

Is that 70% ?

yes. I thin energy needs to be more heavily weighted.

Was just going to say

zas: I need to get some work done before the end of the day. might be best to sleep on this anyway as I can't see the wood for the trees right now.

goddamn docker

maybe valence as well

atj: ok

sorry, it was going well up until that point

np, shit is always expected ;)

I think this could reasonably work as a whiplash detector.

btw, I wonder about rex<->rudi communication, because connecting from rex to rudi using redis-cli doesn't work better

whatever I try it seems keydb cannot answer

so I suspect firewall

that would be my instinct too, but the docker process is listening on *:13062 and I can see traffic in tcpdump, but it's resetting the connection

I'll think about it, might fire up a test container to replicate the issue using nc or something

monkey: I mentioned playlists made from nothing but a seed track. give me a no too obscure recording and I'll make you one -- I'm listening to one right now and its quite listenable.

this offer is good for anyone else too. :)

How about… this? https://musicbrainz.org/recording/2d5ae8c3-0db6...

that looks listenable. :)

Indeed

anyone else? atj zas? gimme a seed recording mbid and I'll make you a playlist of similar tracks. best of it isn't too obscure.

not too obscure? well, try https://musicbrainz.org/recording/4a714ff1-5c44...

that also look acceptable upon inspection

2e2764e4-9b9f-4ff3-816b-122775d97d0e

yes; all tracks are from +/- same period, and genres

please do give me feedback on these once you've listened to them.

!m lucifer

You're doing good work, lucifer!

the playlist has many tracks I already know, but almost none are in my listens, only 2 were unknown to me

looks good, a couple of outliers

thanks monkey, your list is quite the breath of fresh air for what I have been listening to.

I have similar feedback (not really a need to listen to the whole things, knowing pretty much all the songs on there): there is a coherence of time period and genre for the most part. A few outliers (more recent bands, adjacent genres) but for the most part this could be a 70's rock mix

zas: thanks! that feels about right for what our goal here is.

atj: are the outliers jarring to you?

Blur - Song 2, The Cure - Just Like Heaven, Jamiroquai - Virtual Insanity (borderline)

Song 2 is definitely jarring

the rest are excellent

I know almost every song but I don't think many are in my listens

monkey: zas: as far as the "i know this stuff" comment -- when we make an artist radio, we'll pick similar artists and their popular/kinda popular tracks and then get similar tracks from those... those are less guaranteed to be known

I'm curious to see how it does with more obscure stuff. Could you please generate one for https://musicbrainz.org/recording/ce678bd7-687e... ?

It's obscure but a very well defined genre/area

:)

atj: https://labs.api.listenbrainz.org/similar-recor...

depending on how long this takes to generate / where the data is, it'd be neat to have debug output in dataset hoster

oh hey. dataset hoster :)

quite interesting it included "Grover Washington, Jr. - Knucklehead", which is not the same genre but is a song sampled in loads of hip hop

those three tracks are what I would call "well into the weeds", so all good on that front. ?)

!

especially debug info to answer questions like "how many people have listened to this song in our dataset" - to see how obscure monkey's track actually is

is the score out of 100?

alastairp: monkey : no similar recordings found.

obscure stuff has fewer similar tracks.

OK, not super surprising

that may change as we keep tweaking the alg.

atj: no, the higher the score, the better.

atj: it counts the number of co-incidences between the two tracks in user listening sessions.

right

so with MLHD and with more users, obscure tracks will become less so.

yeah, for sure

but if we can get this to work reliably for popular and medium popular stuff, we're good to go.

how about this, shouldn't be too obscure but the genre is... a bit nebulous 59a1ccec-616d-46de-aa46-a160242ee7d5

I don't know FKA twigs but the rest seems good

that must be more obscure as there aren't many tracks

I just discovered Burial because of this work.

bonobo, squarepusher and alt-j in the same list seems pretty logical

Untrue is a good album.

DjRum and Synkro are pretty similar

seems very promising!

I also have a feeling that there is some heavy bias from some specific users that can be felt

yeah. I'm pretty pleased. I'm further down the path towards something usable than I previously thought.

monkey: yes, very much so.

Wet Leg everywhere in all these recommendations when it comes to anything related to rock

I'm guessing that's from an old LastFM user with lots of listens

yeah, I see my tastes all of my listening neighborhood as well.

I've notices Spoon as well appearing in lots of places

but, this playlist patch very specifically does limit the playlist to two tracks by the same artists.

a radio alg would do that.

But that's a problem that will become less pronounced over time

need more data

so, our proclivities will be less visible.

Right

atj: that. the challenge is to make do with what we have now. but I think we're nearly there.

If I'm honest, all these playlists above sound reasonable to me