#metabrainz

Temporarily disabled live indexing to launch a full reindex.

yvanzo: great

I was gonna modify some of the commands in the upgrade steps to use admin/run, but I'm not sure it's working properly (or I don't know how to use it): https://gist.github.com/mwiencek/d3b0f3c1c98311...

I guess it's not really meant for interactive commands. it works if I pipe a query into it. but the "Running 'MAINTENANCE'" bit threw me off

!m yvanzo and bitmap

You're doing good work, yvanzo and bitmap!

eh. idk. "swipe and hold" or whatever

I'm 40yrs old. i like having thing to click

To be fair, us old people are probably not the main target market :D

yvanzo, bitmap: I did some basic testing (tag, rate, untag and make sure it's removed, check area containment shows fine, check withdrawn RGs show again)

Merge some artist credit with Edit AC and ensure it didn't scream

The genre and mood stuff isn't implemented so I can't test that, but I have tested it locally when implementing it and I think it should work. Mostly, at least :D

(I didn't test any other rels than url rels because of the mini rel editor not working in React yet)

The rest of the AC MBID changes aren't implemented either but I understand yvanzo has implemented and tested some of that locally too

reosarevok: I mean yea. but like

some day th target market will also be old :D

By then apps will be for old people and young people will have something else :p

Probably something awful :D

I'm already mad about how bad it will be

Let's yell at some babies, it will be their fault ;P

O’Moin

(still importing releases)

hi, so who should i talk to about something about listens

monkey maybe?

my lucifer

erh hmm*

idk mayhem, anyway

for my page https://beta.listenbrainz.org/user/CatCat/ if you scroll to the bottom

you see CUSIN BONELESS rtracks and then the Bridge City Sinners

well theseare all on the same album, but for some inexplicable reason tthis is saved by being split into two differnt albums. nither of wich match what it's actually is

and i see this for *A Lot* of things in my listen history

like, does the saved album name make NO distinction??

your model is kinda warped if the name of the album of the track being listened is ignored completely in your model

i mena that's literlaly the issue with last.fm

moin yvanzo. Indexing, you mean?

yes

Is releases generally the last very long one? Or is RGs still to come?

No, recordings are longer.

ok I know I'm complaining about this literally on the worst day possible, (scemarelease day) but like i've mentioned this a few tiems before and it was soundly ignored

i'm sick of that

RGs come next

afer the schemarelease this sould be dealt with

They did

Ooh

I meant of what's left to index :)

Ok, so RGs will also take a while I guess

It's alphabetic order

But hopefully it'll finish

CatQuest: LB is not connected to the schema change anyway, it's just the comments might (or might not) get drowned in schema change talk :D

I assume there's a ticket?

reosarevok: i know, comments get drowned

idk about tickets. if i create a ticket with a lin kto my listen history liek this it measn i can't import any more listens becasue the it'l lbe overridden like

but if you can help me wednesday I'll make a ticket

CatQuest: i see, that's probably an issue with mbid mapping. mayhem has planned some changes to the process so that it can identify albums better. but its not there yet and will take some time to land. ofc once its done, we could run it retrospectively.

lucifer: awesome!

yvanzo: reosarevok: in my experience it takes about 9-12 hours to run full indexing on wolf.

lucifer: I thiink it’s taking so long mainly because I did not tune SIR config for a full reindex.

oh i see.

maybe stop and configure and run full index only on the subset of remaining entities?

Moin!

morning

How are things on the schema change front?

It won’t take too much time fortunately, I will let it complete. And then rebuild one index with a better config to see the difference.

makes sense

Is there a planned time to start?

Time for a quick blog post?

mayhem: Not too bad, had some issues spotted and fixed on yesterday. There still are ongoing tests.

mayhem: 17 UTC: https://twitter.com/MusicBrainz/status/15258762...

Ok, great.

Keep me posted and ping me if you need help I'm going to be available to help out today, ready to backup whatever is needed.

I guess that anything that depends on MB will have to be put on hold.

[musicbrainz-server] 14yvanzo opened pull request #2532 (03master…fix-run-output): Fix echoing the command to be run https://github.com/metabrainz/musicbrainz-serve...

bitmap: Fix the output of admin/run ^

bitmap: This is not related to using interactive commands; for example "admin/run tac" works fine.

bitmap: In its current state, it is meant to run in production containers only. It has to be improved for use in other environments because it makes some assumptions (user has to be 'musicbrainz', PATH variable, PERL_CARTON_PATH variable, log directories).

It isn’t logging the input either.

yvanzo: will the DB be in read-only for a while? how long will it be down for totally?

mayhem: I’m not sure about the read-only mode, something has been implemented and used for the last schema change but it is not documented yet.

The DB is expected to be down for about an hour.

bitmap: Reported requested improvements to admin/run for later on at https://tickets.metabrainz.org/browse/MBS-12384

MBS-12384: Make the command logger usable beyond production containers

lucifer: I think last time we took a full downtime for LB, no?

because logins would fail, at very minimum.

alastairp: ping

mayhem: i don't remember if we took a full downtime, i do remember running spark tasks at the same time so at least some part of LB was up.

yes logins would fail, don't think anything else should.

if MB db is going to have a downtime, we don't have a choice anyway. if its read only, LB can remain up

why dont we have a choice?

what if we just disable logins during the downtime?

consul won't let LB come up

that. yes, that will do it.

not entirely ideal, sadly.

then full downtime it is. we should really improve this for next year.

right we can disable logins and make MB dep conditional like TS but not sure if that can be done before today's schema change.

morning

mayhem: hi, here

morning, perfect timing alastairp .

just reading

see our discussion just now.

yes makes sense to improve it

thanks

logins will fail, but we also need to check API auth tokens

which requires the same database

which API auth tokens?

submitting listens

the MB DB will be offline, the LB DB will still be online.

the cluster will remain available afaiu just not the MB db

yes.

hmm, let me read backlog again more carefully

so, auth tokens are stored in LB, so that should be fine.

i can try to whip something out so that we can keep accepting listens

lucifer: that would be ideal. hopefully alastairp can help a bit.

yes, I can help too

cool, i'll follow the same thing we do for TS.

my availability is rather unpredictable today. Everything is priority A on fire today. :)

I'll be around, but can't really help do serious continuous work today. #shortattentionspanday

OK, I understand that the main metabrainz pg cluster will be available, but not the musicbrainz database, is that correct?

correct

OK, now everything makes sense

so we disable login, but if someone is logged in then they're fine, and if they submit listens we can validate against the user table just fine

that is my understanding.

yes, mine too

yup makes sesnse