rdswift: we may release Picard 2.9 alpha version before Christmas
yvanzo
lucifer: Do you remember why we had SIR failing with - 'sessionmaker' object has no attribute 'query' - prior to merging the SQLAlchemy 1.4 PR (that was on dff0a113)?
yvanzo: oh, ok :) These were theoretical PRs, not tested locally - and I never claimed I had :) (hence the AFAICT disclaimers)
But now we have proper indexing tests, I guess I could look into add some tests for those at the very least
s/into add/into adding/
yvanzo
Unfortunately we don’t because these have been added along with this same issue.
reosarevok
Oh, I guess I misunderstood the "this commit had fixed the issue" bit as "we can just fix the issue applying this on the branch" :)
yvanzo
reosarevok: As a workaround it can based on the branch "released-master" (which matches the latest v3.0.1); See PR #157
reosarevok: The fix comes after the migration to SQLAlchemy 1.4 which isn’t running properly for now.
reosarevok
Oh, so it's 1.4 dependent? Ok then
Sophist_UK has quit
yvanzo
I tried cherry-picking at least.
The branch "released-master" is temporary only until the next release whichever new features set or stable SQLAlchemy 1.4 comes first.
reosarevok
How's the 1.4 improvement going? I saw more and more entities are getting back to 3.0.1 speeds
Sophist-UK joined the channel
yvanzo
recording index is the blocking point atm
and the only blocking point for full reindex
I don’t think that live reindex will be an issue but it cannot tested for perfs without recording
that's about it :)
lucifer
yvanzo, for live index i don't think recording will do OOM. could try testing it.
i am not sure why it OOMs for recordings but simplest explanation would be that it has the largest amount of data to process. for live reindex, the number of entities being processed at a given time are low so might not be an issue.
yvanzo
lucifer: yes but it requires a full index to test it for real :)
lucifer
ah ok, makes sense
yvanzo
I’m not worrying about it either
lucifer: maybe we can add more parameters to config.ini if that can help with fine tuning it?
if there is a need for specific settings not to OOM on recording for example
lucifer
yes makes sense i think. maybe lower number of processes or batch size for only specific cores: recording for now.
yvanzo
reosarevok: Getting a better picture of dependencies for new search features :)
Maxr1998 joined the channel
Maxr1998_ has quit
jivte joined the channel
jivte
lucifer: Hey fixed the error and pushed the code
lucifer
jivte: thanks, running actions again. if everything passes, i'll merge it.
jivte
okk and thanks so much for being so helping me out
I just want you to be aware of it. However, this doesn't mean that we do not want to work on the MusicBrainz android app. I love it!
jasje has quit
We just need more focused devs. Feel free to work on either/both the apps in future as per your convenience! Thank you so much!
Once your PR for MB android is merged by me, let me know what you would like to work on next. It'll be great if we have a discussion on your interests for the apps
jasje joined the channel
jasje
akshaaatt: Yes i know i read the blogposts on newsbrainz
i would like to work on listenBrainz primarily but ofcourse work on anything necessary for musicBrainz
Mainly will do anything that alleviates some pressure off of you guys
jasje has quit
jivte
lucifer: Hey I would like to work on LB-1046 ticket and if I can work can u please explain the ticket a little more
Lotheric: Thanks. Would be easier if it were on the 'main' site. Will work with that one unless I find another.
Ah, seems to be accessible from the main URL too. Thanks Lotheric.
lucifer
jivte: i need to discuss how to implement that with monkey. also how the ui should work will need to be decided with monkey and aerozol so i suggest to leave it for now. (monkey is not around today)
Lotheric
adhawkins, yeah beta and main share the same database
this is the complete list of fields available. some of these are not used by the `_get_playable_recommendations_list` function.
to fix the issue, look into other uses of `load_recordings_from_mbids` in the codebase and then add the fields which are missing in `_get_playable_recommendations_list` by taking help from how its done in other places
jivte
okk understood will try and then inform you
lucifer
👍
aerozol, reosarevok: does showing release group cover art as a fallback for listens solve LB-1119?
The ticket is right that the best option would be to follow the relationship
But it might be fine to just fall back to the RG tbh
Especially since soon bitmap will implement the alternative tracklists and pseudos will go away, right bitmap? :D
lucifer
ah cool. getting that link and following it would be too much work for not much benefit i think.
especially because currently we don't have a way to detect whether a release is psuedo release in LB.
reosarevok
yvanzo: I rebased those two SIR PRs to be on top of released-master
What should be the next step?
rdswift
<zas> "rdswift: we may release Picard 2.9 alpha version before Christmas" Okay, thanks. I'll be away for a week or two around Christmas so I'll merge all 2.9 documentation changes into a single branch and keep it up-to-date so that there will be only one merge required in case I'm away when it happens.
yvanzo
reosarevok: Testing it
outsidecontext
rdswift: thanks
rdswift
Actually, two merges because the second one will be to update the translation files. ;-)
Will the 2.9 release be tagged as an alpha or beta?
So I get it right in the config file.
yvanzo
lucifer: I silenced the alert for a day
lucifer
yvanzo: thanks!
yvanzo: the recording index has currently been running fine on wolf with 4gb shared buffers. ofc taking longer. can you share the 1) vm specs 2) postgresql.conf 3) oom logs both pg and linux syslogs of the vm that crashes with recording core.
rdswift
Sorry, I just re-read zas' note. Alpha it is.
lucifer
i want to try and reproduce the error on wolf so that we can know what exactly causes the crash.
yvanzo
lucifer: 3½) PG logs are under ~musicbrainz/log-dc96c30-master/*.db.stderr.log already
lucifer: 1) which specs/commands do you need exactly?
KassOtsimine
reosarevok> Especially since soon bitmap will implement the alternative tracklists and pseudos will go away, right bitmap? :D riiiight bitmap??? 🥺
lucifer
yvanzo: yes but those are from wolf where the index completes. i am talking about the db logs when the indexing crashes. re specs, i think `lscpu` and `free` should suffice.
yvanzo
Oops sorry!
lucifer: for which branch/commit of sir?
lucifer
yvanzo: current master probably but any commit where recording core crashes would work
I’m be looking for logs now (if these are still available)
jivte has quit
Aswin joined the channel
Aswin has quit
lucifer: full logs (sir, pg, dmesg) for last OOM on building recording index under ~musicbrainz/log-recording-14db7a1-raiseload-rebased-on-dc96c30-master/
(on wolf)
lucifer
thanks yvanzo!
yvanzo: on this vm can you trigger a recording index with cores configured to 2 (it has 4. so half)?