we don't actually link the user's apple account with LB because that needs us to load the apple music player script.
2023-07-21 20246, 2023
lucifer
there is no proper oauth api to do that.
2023-07-21 20209, 2023
lucifer
i was wondering if on clicking this button, we could load the music player dynamically
2023-07-21 20236, 2023
lucifer
and then have it perform the oauth, save the token back to LB via an api call.
2023-07-21 20204, 2023
monkey
Yes, loading the library dynamically shouldn't be an issue
2023-07-21 20208, 2023
lucifer
currently, this happens when the user tries to play songs on LB frontend with the apple music player.
2023-07-21 20235, 2023
lucifer
i think the workflow outlined above is better, thoughts?
2023-07-21 20218, 2023
monkey
Yes, definitely a better idea IMO to try to do as much setup in one go when the user enables the feature
2023-07-21 20232, 2023
lucifer
cool, makes sense. one more question.
2023-07-21 20218, 2023
lucifer
when the linking expires because of the tokens expiring, we'll probably get some 403s. to have the user relink it we just send the user back to that settings page?
2023-07-21 20224, 2023
lucifer
and ask them to relink?
2023-07-21 20224, 2023
monkey
Was that the issue that the token expires really quickly?
2023-07-21 20255, 2023
lucifer
i think in the best case it will last 6 months.
2023-07-21 20204, 2023
monkey
Can we detect this specific case (organic expiration of token)? If so we should maybe try to handle it in BrainzPlayer when we encounter a 403. But as a first version, I think it'll be simpler to link to the settings page for the user to reauth
2023-07-21 20223, 2023
lucifer
makes sense
2023-07-21 20229, 2023
lucifer
thanks!
2023-07-21 20219, 2023
monkey
Of course
2023-07-21 20236, 2023
lucifer
pixelpenguin: hey!
2023-07-21 20220, 2023
monkey
Generally, I've had a few ideas brewing for improving how we manage third-party music playing settings.
2023-07-21 20220, 2023
monkey
I think a new BrainzPlayer settings page will be necessary, and will likely replace the current music service page
2023-07-21 20258, 2023
lucifer
monkey: yup indeed, that would make sense.
2023-07-21 20233, 2023
monkey
Do you reckon we could save separate tokens for reading data and for playing music? For example for Spotify we could allow users to link their account to read history, and separately have another token for BrainzPlayer.
2023-07-21 20248, 2023
monkey
Would probably simplify some of the scopes mess we have to deal with
2023-07-21 20209, 2023
lucifer
that might be possible but I am unsure. i think it would vary service to service.
2023-07-21 20218, 2023
monkey
Right.
2023-07-21 20234, 2023
lucifer
one possibility could be to create multiple oauth apps, one for BP and one for LB main.
2023-07-21 20244, 2023
monkey
Right, I guess that makes sense
2023-07-21 20235, 2023
monkey
PossiblyNot sure it would be necessary though. I'll keep brainstorming.
2023-07-21 20213, 2023
lucifer
yup sure
2023-07-21 20234, 2023
monkey
Oh, lucifer, one more thing if you've never done it: to load a script dynamically, you can write some JS to inject a resource into the page (like you would with HTML) and it will load the library
2023-07-21 20216, 2023
lucifer
monkey: yes makes sense, i did see the utils functions to do it with react.
2023-07-21 20235, 2023
lucifer
on this page, might need to write more JS from scratch because its not react-ified yet.
atj_: not yet, just so I could measure how each config change affects things, but planning on it today
2023-07-21 20235, 2023
atj
bitmap: do you use `wal_sync_method = 'fdatasync'` by default?
2023-07-21 20227, 2023
bitmap
yes, it's set to fdatasync
2023-07-21 20236, 2023
atj
oh, it's the default on Linux
2023-07-21 20226, 2023
pixelpenguin
<lucifer> "pixelpenguin: hey!" <- heyy, what's up
2023-07-21 20205, 2023
jasje
akshaaatt: update dev branch?
2023-07-21 20209, 2023
jasje
with my pr
2023-07-21 20213, 2023
jasje
work all good
2023-07-21 20227, 2023
jasje
ship it
2023-07-21 20209, 2023
bitmap
atj_: setting full_page_writes=off doesn't seem to affect the benchmark at all, probably because the test doesn't write enough data or run long enough to trigger an automatic checkpoint
2023-07-21 20251, 2023
bitmap
were there any other settings I should tweak/? jimmy has the same amount of RAM as pink/floyd right?
2023-07-21 20242, 2023
atj
bitmap: you can disable wal_init_zero
2023-07-21 20221, 2023
atj
and wal_recycle
2023-07-21 20247, 2023
bitmap
ok
2023-07-21 20212, 2023
atj
the other tweaks are at the ZFS level so we'll need to arrange a time to collaborate on that
2023-07-21 20252, 2023
atj
but performance seems reasonable at this stage?
2023-07-21 20204, 2023
bitmap
yeah
2023-07-21 20226, 2023
bitmap
I could start sending standby traffic to there (currently handled by pink)
2023-07-21 20210, 2023
atj
best hold off on that for now
2023-07-21 20255, 2023
atj
I think some ZFS tweaks may have a decent impact on performance, but that will require creating a new FS, stopping the container and moving the data across
2023-07-21 20224, 2023
atj
basically we need to create a dedicated FS for PG data and WAL
2023-07-21 20232, 2023
rdswift has quit
2023-07-21 20254, 2023
rdswift joined the channel
2023-07-21 20247, 2023
lucifer
pixelpenguin: Hi! let's discuss about setting up select options in Dataset Hoster UI. so like we currently have input fields for text and datetime, we want to add a field for select input that should be prepoulated with various options. these options can be hardcoded as an enum in the start.
2023-07-21 20252, 2023
lucifer
in the final format, we want the options in the enum to be dynamic if possible (for example load from redis or database)
but either way you can start implementing the part with static options first and then we can look into dynamic options.
2023-07-21 20253, 2023
lucifer
for the partial query ticket, i will have a discussion with mayhem and then you should work on it next.
2023-07-21 20248, 2023
pixelpenguin
Thanks for the detailed explanation of the implementation, I agree with it.
2023-07-21 20204, 2023
pixelpenguin
I can get started on it
2023-07-21 20241, 2023
pixelpenguin
Is there anything else I should look out for?
2023-07-21 20243, 2023
pixelpenguin
Or any other task in the current PR which I should work on first?
2023-07-21 20239, 2023
lucifer
pixelpenguin: i had looked at the current PR, it lgtm. i have not merged it yet just in case some bug comes up and we need to do a new bugfix release of DSH. the current PR introduces multiple incompatible changes and that would make it harder to do so.
2023-07-21 20200, 2023
lucifer
so we should probably merge it all at once.
2023-07-21 20238, 2023
lucifer
so no other changes required on the existing work but add the changes for these remaining 2 tickets to that PR as well.
2023-07-21 20216, 2023
bitmap
atj: ok, let me know if you create the new FS, I can help move the data over. though I'm not sure how well I can measure the impact of those changes with pgbench alone. I don't see any disk ops stats for jimmy in grafana
