hmm, I think the problem is that the LB web container didn't reload uwsgi when I put the old pgbouncer container into maintenance mode. this template block in /etc/consul-template-uwsgi.conf looks suspect since it has no command/exec setting:... (full message at <https://matrix.chatbrainz.org/_matrix/media/v3/...>)
lucifer: is my understanding of the consul-template config correct? if so, can it be changed to reload uwsgi when config.py changes?
that seems problematic since the workers have to be restarted if the config changes, right?
zas: hey, can you configure gmail to allow exim relay on isaac? (logging off now but I can move them tomorrow)
lucifer[m]
yes, i think so. i'll discuss with monkey on the rationale.
aerozol[m]
mayhem: 🥳 v
Aerozol[m]1 joined the channel
Aerozol[m]1 uploaded an image: (17KiB) < https://matrix.chatbrainz.org/_matrix/media/v3/download/chatbrainz.org/zcAxxQAkKnBzmftKQlQTuboT/image.png >
pite has quit
dabeglavins3 has quit
dabeglavins3 joined the channel
BrainzGit
[bookbrainz-site] 14shalini-bhandari opened pull request #1154 (03master…replace-twitter): fix: Replaced Twitter logo and link with Mastodon and Bluesky https://github.com/metabrainz/bookbrainz-site/p...
Kladky joined the channel
allen joined the channel
adhawkins has quit
adhawkins joined the channel
holycow23[m] joined the channel
holycow23[m]
I wanted to work on the user-stats is there any method by which I can get some stats in my local spark so I could test the generation and retrieval of new stats?
elkiensad[m] joined the channel
elkiensad[m]
<Aerozol[m]1> "image.png" <- yeah the lists are getting better
mayhem[m]
elkiensad[m]: and that is only because new users joining. we haven't made any improvements.
elkiensad[m]
The number of listens of persons I have made a create account on ListenBrainz has reached around ~1012364
2 of them are in the top 2 rows of my similarity chart with 24% and 26%
and i can tell generated lists are mostly copying their tastes at this moment
reosarevok[m]
Oh, so it's an Estonian/Finnish paradise
Get music recommendations from your friends without actually having to talk to them!
lucifer: would you do me a favor and see why daily jams failed last night?
lucifer[m]
Not enough listens in playlist, [@mayhem:chatbrainz.org](https://matrix.to/#/@mayhem:chatbrainz.org)
mayhem[m]
ah, yes. I could've foreseen that. ok, thanks for checking.
kellnerd[m] joined the channel
kellnerd[m]
I couldn't believe there is no ticket for a faster MB API endpoint to lookup multiple URLs and map them to MBIDs (or at least I couldn't find it), so I finally created MBS-13943.
because for the new fuzzy mapping server, making a multi-process version is going to be a serious pain, but deploying with python 13 with no GIL would be possible
nice to get very clear answers. :)
zas[m]
but 3.13 no GIL is still experimental, so perhaps a bit too soon
mayhem[m]
yeah, a bit worried about that.
but making a multi-process version of this would mean using shared memory, which might be impossible to accomplish.
outsidecontext[m
I'm really curious about how this performs in various scenarios, but I haven't tested it yet myself. But according to the release notes single threaded performance seems to be rather highly impacted by this, and there is a lot they want to improve for 3.14.
mayhem: Yes. This is roughly how the discussion about using processes in Picard instead of threads because of GIL came to a halt
zas: we use some serious magic libs to make this matching happen (nmslib) and it would need to support buffers being passed in so that everything could be running on shared memory.
this code is amazing, unsupported and doesn't even have support for being serialized out to disk.
its its hands down better and faster than anything else I've ever used.
bitmap: MBS-13943 seems sensible, I guess - do you have any suggestions about how many it should allow in one go, or any reasons why it'll be a mess? :)
the default max limit of 100 seems sensible to me. and we def. have a ticket for looking up multiple entities at once, just not one for URLs specifically I guess
kellnerd[m]
There is a ticket for lookups by multiple MBIDs, but nothing related to browse requests (of which URL is a special case where it is beneficial to have this).
oh I'm not sure why the docs classify that as a browse request, it only returns a single entity
minimal joined the channel
pite joined the channel
zas: looks like exim is working on isaac, thanks
julian45[m]
yvanzo: FYI, Jira still shows a banner (at least on the main page) for yesterday's maintenance
zas[m]
julian45: I removed the banner
julian45[m]
Speaking of Jira, are we running Server or Data Center? Just taking a look at release notes for a couple versions
* we running/licensed, * for Server or
mayhem[m]
I forget where we are at with that -- yvanzo would know.
julian45[m]
Roger that. If it happens to be Data Center already, the current installed version (9.12.18, latest of 9.12.x LTS series) hit EoS almost exactly a year ago and will hit full EoL later this year. Atlassian has put out a 10.x series, with 10.3 being the LTS release; summary of changes across versions here: https://confluence.atlassian.com/adminjiraserve...
...and if we're on Server, 9.12 is the last Jira version to support that license type, which is fun.
Theoretically they have a planner/health checker/etc. in the Jira admin interface already, but I don't have that level of access in our Jira instance
SothoTalKer__ has quit
SothoTalKer joined the channel
yvanzo[m]
We have a Data Center instance.
The EoL for the current 9.12.x (which is an LTS) is 29 November 2025.
We should indeed upgrade to the next LTS 10.3.x (released on 05 Dec 2024) before that.
Ideally, we should also have this handled through Ansible.
