lucifer: now disable the mbid_mapper via labs api. load now finally dropping on typesense
2025-12-09 34301, 2025
mayhem[m]
s/disable/disabled/, s/mbid_mapper/mbid\_mapper/
2025-12-09 34346, 2025
mayhem[m]
well, we knew that eventually this would go bad. here we are. 🤪🤪
2025-12-09 34341, 2025
mayhem[m]
ok, things seem stable, I'm off to bed. lets see who screams in the morning.
2025-12-09 34302, 2025
bitmap[m]
I'm back, things are loading fine for me here
2025-12-09 34308, 2025
julian45[m]
both MB and LB load for me
2025-12-09 34328, 2025
Pokey has quit
2025-12-09 34305, 2025
v6lur has quit
2025-12-09 34313, 2025
Pokey joined the channel
2025-12-09 34356, 2025
Maxr1998 joined the channel
2025-12-09 34313, 2025
Maxr1998_ has quit
2025-12-09 34331, 2025
ApeKattQuest joined the channel
2025-12-09 34325, 2025
MonkeyPython has quit
2025-12-09 34337, 2025
Pokey has quit
2025-12-09 34300, 2025
Pokey joined the channel
2025-12-09 34326, 2025
julian45[m]
[@reosarevok:chatbrainz.org](https://matrix.to/#/@reosarevok:chatbrainz.org): ignore until work hours in your timezone, but: do you think there’s any chance that MB could cache its own copies of transcluded docs pages rather than (seemingly) pulling them live from the wiki? one thing i noticed during wiki maintenance work was that during periods when mediawiki was unavailable, docs pages on MB would show a “not found” error.
2025-12-09 34326, 2025
julian45[m]
if it’s not possible to cache copies directly, then i’d at least like to know if the error page can safely be changed to something reflecting that the link used is valid and leads to content under normal circumstances, but happens to be unavailable at that particular moment
2025-12-09 34332, 2025
julian45[m]
also, out of curiosity: do MB mirror instances also pull docs from our wiki? if so, while that kind of makes sense to me from one standpoint (namely, mirrors always show up-to-date docs), it also seems to me like a single point of failure and/or a risk for a matt mullenweg type situation, where we hold all the metaphorical cards for all instances of MB even though anyone can nominally run their own copy
href="//wiki.musicbrainz.org/specific%20topic… topics, methods, or questions you study</a>. I use Zotero to organize literature, manage citations, and collaborate on research projects.<br /></bdi></p>`
NOTICE: MBS-14178 (https://tickets.metabrainz.org/browse/MBS-14178): Indicate the edited release on recording and RG edits entered from the release editor
NOTICE: MBS-14210 (https://tickets.metabrainz.org/browse/MBS-14210): Add mixesdb.com to the other databases whitelist
2025-12-09 34321, 2025
mayhem[m]
zas: any idea why I might be having issues connecting to our sites? happens on different browsers, but not when I use a VPN. no problems with other sites... very odd
NOTICE: MBS-14178 (https://tickets.metabrainz.org/browse/MBS-14178): Indicate the edited release on recording and RG edits entered from the release editor
2025-12-09 34319, 2025
reosarevok[m]
mayhem: are you running a malicious bot? 😝
2025-12-09 34354, 2025
mayhem[m]
malicious brot, maybe....
2025-12-09 34314, 2025
rozlav has quit
2025-12-09 34300, 2025
reosarevok[m]
Quickly updating the prod servers with a small patch so beta edits don't crash after beta release
2025-12-09 34319, 2025
mayhem[m]
lucifer: let me know when you\'re around, ok?
2025-12-09 34329, 2025
lucifer[m]
mayhem: i am around
2025-12-09 34357, 2025
mayhem[m]
so, the traffic loading typesense down comes from the labs api mbid_mapping endpoint.
2025-12-09 34321, 2025
lucifer[m]
do we know its our mapper or external traffic?
2025-12-09 34330, 2025
mayhem[m]
not sure if the metadata api endoint has much traffic, but putting a login required there is a good idea.
2025-12-09 34341, 2025
lucifer[m]
i can look at gateway logs to try to determine if not. iirc we don't apply ratelimits to labs.api.
2025-12-09 34355, 2025
mayhem[m]
our internal traffic, at least from our own mapper, uses typesense directly.
2025-12-09 34316, 2025
lucifer[m]
we have multiple metadata endpoints. one in LB directly, one through labs.api.
2025-12-09 34320, 2025
mayhem[m]
I had the labs api endpoint return {} last night an none of our systems were affected. all cron jobs (mapping) returned to normla.
2025-12-09 34332, 2025
mayhem[m]
I disabled both of them
2025-12-09 34344, 2025
lucifer[m]
makes sense
2025-12-09 34306, 2025
mayhem[m]
I think we should: remove the labs api mapping, release mapping and mapping explore endpoints. and make the /metadata/lookup endpoint auth required.
2025-12-09 34307, 2025
lucifer[m]
i think we can just put them behind a very conservative rate limit for now. i should have a PR ready soon.
2025-12-09 34325, 2025
mayhem[m]
and given that no one has screamed about this.... I think its AI bots hitting it.
2025-12-09 34327, 2025
lucifer[m]
in favor of removing labs.api endpoints for which we have alternatives.
2025-12-09 34348, 2025
lucifer[m]
for /metadata apis there are genuine consumers.
2025-12-09 34301, 2025
mayhem[m]
ok, I'll do that then, ok with auth on /metadata/lookup ?
2025-12-09 34305, 2025
lucifer[m]
{} is a valid response which means no match found.
2025-12-09 34337, 2025
lucifer[m]
i think adding auth is fine but also let's do a blog post about it.
2025-12-09 34338, 2025
mayhem[m]
lucifer[m]: yep. that is all I had the mental capacity for yesterday. 🤪
2025-12-09 34352, 2025
mayhem[m]
ok, I'll get moving.
2025-12-09 34358, 2025
lucifer[m]
i had returnerd internal server errors in a similar patch earlier.
2025-12-09 34307, 2025
lucifer[m]
so makes sense.
2025-12-09 34341, 2025
reosarevok[m]
Prod should be updated - you should see no visible changes
