unless the date range fits the recording's release date :D
Scruloose: ahaha, but only in a non-coder way :D
Scruloose
Heh! What have I got myself into
CatQuest
hahahaha
welcome to MusicBrainz <3 :D
Scruloose
I think I'm going to ignore the release date thing. The idea of my plugin is to simplify the on-disk directory sorting, not to achieve perfect granularity
CatQuest
sounds fine
Scruloose
so chuck anything with an end date, prioritize 'Primary' if it exists, grab first locale en (or en_<region>), and if none of those exist fall back on top-level "artist" dict values
CatQuest
I mean. if oyur user wants english
i'mnorwegian o probably woudl wnat norsk names? :D
CatQuest is helpful :P
norsk is another can of worms. so i'll leave yo uat that :D
Scruloose
Oh! Yes definitely. I'm also planning to learn how to get user's locale out of beets config if they've set it, and only use en as default
CatQuest
i hope it has a en-
erh
i hope it has a en_GB option as well. if i cna i use en_gb over en_US
if not.. miht wanna ask for it upstream.
hmm.. wasn't it Freso working on/with beets.. no that was whipper wanst it
CatQuest really goes
Scruloose
Thanks for the info!
chaban joined the channel
yvanzo joined the channel
adhawkins
ruaok: Can you point me at some documentation regarding tuning postgres' memory usage for the musicbrainz server? About to increase the RAM on my VM, so want to make the appropriate change there.
ruaok
not sure I have a good doc to give you, but the setting that needs adjusting is shared_buffers, which should be set to 1/4 - 1/3 of available ram if you can give it that.
if you're using the VM, then most of the settings are tuned well enough that you only need to tweak shared_buffers.
We're planning to move some docs to github eventually, but for now they're all wiki-based
revi
well I don't know about deep-inner-workings of this project but sounds good to me
on Wikimedia we limit people to 2 rename requests
(for third and more you need to have reallllllllllllly convincing reason other than "I changed my nick)
reosarevok
Seems sensible :)
I'll transclude the change then. Thanks!
Scruloose
Is it safe to assume that an artist dict returned by musicbrainzngs.get_artist_by_id will *always* have both a name and a sort-name in the top-level dict?
Or, to put basically the same question another way, are bolded fields in the flowchart ruaok linked mandatory?
reosarevok
Yes
Scruloose
Yay! I found something that's as simple as I thought!
reosarevok
bitmap, yvanzo: I know I've asked before but I forget - do we know why our Selenium tests have random failures so often?
ruaok
Scruloose: yes, seems so.
yvanzo
reosarevok: race conditions, but I don't recall exactly.
KassOtsimine has quit
Cyna
bitmap: so are you considering having a wrapper for mocha to html test ?
currently the supported export types are JUnit, Pytest and Js mocha
for our case js mocha would be the better option
reosarevok: could you link me the doc where instruction for running test for mb-server is present please ?
I tried finding it but failed
bitmap
no plans atm, what we have is working for now so I'm not too interested in changing it yet
[listenbrainz-server] paramsingh merged pull request #635 (master…param/revert-username-changes): Revert "Update the musicbrainz ID of users if it has changed" https://github.com/metabrainz/listenbrainz-serv...
my life is returning to normal finally (mom & bro are on the way back to the US) and after moving the MSB mapping two steps forward, I'm going to buy us a cluster to play with.
iliekcomputers
kinda same with my life.
ruaok
\ø/
I suppose if we have an always on cluster we should aim to keep it in sync at all times.
also, I wonder if we can start with four nodes and not 8.
because I really dont understand out load demands at all.
our
iliekcomputers
for stats calculation, 4 would be more than enough imo
i'm not sure about the recommendation stuff.
ruaok
yeah, this has to work for stats and recommendation.
might be enough, not sure.
iliekcomputers
the hetzner vms did stats calculation on the MLHD very fast.
I think getting the cluster right / making it easy to add or delete nodes is really important here.
ruaok
but given that I am thinking of getting machines that are connected via virtual LAN, it gives me much more flexibility to add/remove nodes.
yes, so the first thing I need to do is get networking right (firewalls, not exposing ports, etc) before we ever load data on it.
iliekcomputers
right.
ruaok
and once that is stable, then we coordinate adding/removing nodes in a careful fashion. since we saw that randomly adding/dropping nodes causes havoc.
not that all havoc is bad, either, but in this case...
iliekcomputers
hmm. setting up the cluster is a blocker for a few PRs for now for me.
will we have consul on the cluster? not sure if too early to ask?
for configurations etc
chaban joined the channel
ruaok
the cluster is substantially going to be the same as it is now. just on proper dedicated hardware.
yvanzo
bitmap, reosarevok: about the react-table thing, both issues have been reported upstream and have open PR in progress.
ruaok
so, no consul.
iliekcomputers
ah
bitmap
yvanzo: great
iliekcomputers
ok.
right, so I can test stuff in the leader cluster and it'd be basically the same.
cool.
ruaok
another thought... when the new cluster comes online, can we add the new nodes to the existing nodes, have them propagate and then take out the old nodes?
yes, exactly.
and perhaps we can migrate the cluster to the new machines, without having to start over.
yvanzo
Just found a new bug with UserAccountTabs which expects EditorT but likely receives incomplete instance when visiting user collections.
iliekcomputers
I don't have enough context onn how the data storage in hdfs etc works to answer that. but i can look.
we made starting over really easy initially, so it shouldn't be much of a problem though.
unless something has changed in the past few weeks.
ruaok
not really, no. starting over should be easy.
but I am wondering if trying to migrate the cluster allows us to learn how to add / remove nodes without breaking everything
also, you only used a partial data set for the MLHD, right?
my next question: how much disk should we get?
iliekcomputers
what do we plan to do with the cluster?
Is mlhd still in the plan, what with the inaccuracy of mbids
ruaok
MHLD, user stats, collab filtering. for starters.
iliekcomputers
I think if we provide a sorta inaccurate dataset with some gamification in MusicBrainz, it might be okay.
ruaok
and I would like to run the whole MHLD to see how that comes out. and then put the dataset up and let everyone poke at it. perhaps its useful, perhaps we can identify problems and maybe even fix them.
