#metabrainz

[mb-solr] 14yvanzo merged pull request #58 (03master…solr-9.6.1): Upgrade Solr version to 9.6.1 and other dependencies https://github.com/metabrainz/mb-solr/pull/58

kellnerd: if I throw https://github.com/metabrainz/musicbrainz-serve... on test wanna test it a bit with harmony?

[mb-solr] 14yvanzo merged pull request #59 (03master…solr-9.7.0): Upgrade Solr version to 9.7.0 and other dependencies https://github.com/metabrainz/mb-solr/pull/59

zas, yvanzo: do I remember correctly that at least one of you has gone to the BCN summit by train in the past? Considering doing that since it seems Brussels - Lyon - Barcelona is fairly trivial with just one change, any bad experiences?

yvanzo: btw, I looked into the mbdata update - it's fairly automatic based on the sql files, so I will make sure to get that ready once we have all schema changes merged, no worries in that regard :)

No, just use the Renfe as much as possible, their trains start in France.

We even happened to book it together.

Renfe rather than TGV-Inoui then? Surprised to hear a French person praise Spanish trains, are the French ones that bad :D

Renfe reduced the number of trains they operate from Lyon (but I think that's still ok from Paris). So check prices for Brussels -> Paris Gare du Nord -> Paris Gare de Lyon -> Barcelona.

Yes, Renfe if possible

TGV Inoui = Ryanair of the train

reosarevok: It is fairly automatic to have issues too. Plus, mbdata is needed to build sir on top of it, which should be delivered at the same time as the schema change.

I would prefer to take SNCF over renfe... #perspectives

Sadly it seems both of them require disassembling to bring a bike, otherwise I'd be really tempted

Well, Renfe trains in France are actually cheaper and better (and there're still TGVs...)

(tempted to bring mine I mean...)

yvanzo: sure, I'm not saying I'll do it after the release :) Just once the tickets are merged

I can point it to the schema change branch rather than master and let it do its thing, and have time to fix any issues that come up

Half of the stuff is merged already :)

Ah yes, bikes in trains, they really want to save the planet, but they still make bikers life complicated when it comes to take trains or planes...

reosarevok: you have a direct Brussels -> Lyon (which is very fast, as it skips Paris)

Yes, that's what I had seen, and why I mentioned Lyon specificall

My understanding is "always avoid Paris if you can" applies to everything in life, probably including trains :p

(although I guess it's getting better maybe? I keep hearing surprisingly positive things about Paris recently)

(is it a trap?)

but beware of railway stations: in Lyon, there are 3: Part-Dieu (center), Perrache (center too), and Saint-Exupery (airport). If a train arrive at Part-Dieu, and another leave from Perrache or Saint-Exupery it is more expensive and more complicated

Yes, avoid Paris :)

For Spain, from Belgium, that will be mainly Part-Dieu, but always double-check that. Same for Paris (if you come from Brussels you arrive at Gare du Nord, and you have to go to Gare de Lyon for Lyon, you need 1h at least (and hope for the best)...)

kellnerd: I'm pushing it to test anyway, can't hurt - feel free to test it!

Ah yeah, that is annoying re: changing stations (esp. since I cannot just cycle from one to the other) so direct to Lyon would seem even better given that

Thanks! :)

[mb-solr] 14yvanzo merged pull request #60 (03master…SEARCH-732): SEARCH-732: Downgrade JFlex to 1.6.1 https://github.com/metabrainz/mb-solr/pull/60

bitmap, reosarevok: I updated mbsssss to v-2025-04-23 while rebasing https://github.com/metabrainz/mb-solr/pull/50 now pending manual tests. Help would be welcome.

bitmap: CI failures come from the `master` branch, likely related to the previously merged pull requests.

RAMDirectory seems to be deprecated

But we don't specifically indicate a directoryFactory...

and mbsssssssssssss uses NRTCachingDirectoryFactory

lucifer: halp? :)

can you point me to the full stack trace?

Search for RAMDirectory

reosarevok: it seems to be an issue in solr-9.7.0, will look into reproducing it locally and fixing it.

Thanks!

<reosarevok[m]> "kellnerd: I'm pushing it to test..." <- Thank you, I will test it once I'm back home. I haven't expected bitmap to implement it within a few hours 😲

What can I say, dreams do come true

[mb-solr] 14yvanzo opened pull request #61 (03master…solrcloud9backups): Add helper scripts to handle Solr backup archives https://github.com/metabrainz/mb-solr/pull/61

Maybe I should declare more of my tickets to be crucial for Harmony if I want them to be prioritized 😁

That might be good for Harmony but not sure if it would be good for harmony

Now I can probably put my idea for a backward compatible API URL change back into the drawer until the move to api.musicbrainz.org

[@reosarevok:chatbrainz.org](https://matrix.to/#/@reosarevok:chatbrainz.org) I was able to fix that error but then it reveals another with setting up cores for tests. Trying to fix that now.

yvanzo: am I right that to make releases searchable by medium MBID, that'd need something like `F("mediumid", "mediums.gid"),` inside `SearchRelease`?

[sir] 14reosarevok opened pull request #160 (03master…SEARCH-738): [WIP] SEARCH-738 / SEARCH-739 (indexer): Support and search by medium MBIDs https://github.com/metabrainz/sir/pull/160

Well, I'll draft some stuff and test it next week :)

lucifer, yvanzo : I'm quite confused rn - where does the test data for mb-solr tests come from?

It's not in the repo AFAICT, but we expect specific MBIDs and stuff so clearly there's some consistent source

But the only source I see for example for this MBID in the MB repos is that same set of tests: https://github.com/search?q=org%3Ametabrainz%20...

reosarevok: https://github.com/metabrainz/mb-solr/tree/mast...

iirc the {core}.xml file is the data indexed into the core, and the {core}-list.xml and {core}-list.json are the files compared to the output.

Oh.

Ok, I thought this would be somehow generated from DB data

Thanks

[musicbrainz-server] 14mwiencek opened pull request #3527 (03schema-change-2025-q2…mbs-13464): MBS-13464: Replace `sort_character` with full `name` for release/release group sorting https://github.com/metabrainz/musicbrainz-serve...

[musicbrainz-server] 14mwiencek opened pull request #3528 (03schema-change-2025-q2…mbs-13966): MBS-13966: Recalculate release group first release dates https://github.com/metabrainz/musicbrainz-serve...

hi bitmap , did some testing and further digging into the issues you were seeing with kanidm and google. still a little hazy due to webs of dependencies, but at this point, the conclusion i've arrived at is that the root cause is a by-design choice in an upstream dependency :/ it makes sense in most contexts, but not in this one of course. thanks google!

on the other hand, the 2nd-line choice for identity provider that was identified in the SSO doc (authentik) should have few to no foreseeable issues; in fact, authentik integration w/ google workspace is supposed to be regularly tested by the dev team (and that's just on the auth front; with the enterprise subscription, that'd enable automatic provisioning/deprovisioning in google from authentik!)

so, i'm going to try standing authentik up and see if that's more viable for us. the main downsides to that i anticipate are that

- it comes with more movable/configurable parts, so will take a bit of effort to slim down to the relative simplicity of kanidm

- it's more resource-intensive than kanidm would be + relies on a bit of postgres (where kanidm doesn't)

hopefully will have good news to share soon!

<julian45[m]> "hi bitmap , did some testing and..." <- Sad that it didn't work out here, but live and learn I guess

Hopefully authentik works out better!

Jade[m]: same here. kanidm is pretty respectable as is, and i was looking forward to the features it has but other IdPs don't. the problem in this case was the python module used to handle OIDC in the OIDC-SAML proxy that was doing the legwork with google

won't put anything except the saml sub as the subject in the resulting saml claim

julian45: thanks for the update, that's a shame about kandim, but I'm happy to help test authentik when it's running (hope that's not too much of a hassle with all the moving parts)

thanks again!

shouldn't be much of a hassle overall; out-of-box config of moving parts is decent, it's just not what is desirable

<julian45[m]> "won't put anything except the..." <- I would wonder if that was worth patching, if it's a single issue\

But then I gues you've got to upstream it

I'm not the one doing the work here 😅

yeah that's the rub

update: now cooking w/ gas on my test user, just need to refine factor options in authentik

* test user with authentik, just

@bitmap: selda exhibits a weird behavior right now, do you know what's happening? keydb is under heavy load

zas: weird, not sure why keydb would be acting up, I think it's only used by test.mb

are there any stats for hits to test.mb?

all of the keydb queries look like they're from normal website traffic, but there is definitely a lot of queries