Will it be possible to make the links in the edit note field (or show them next to the textbox) clickable somehow? Maybe utilizing the url detection from the url archiver?
Usecase: Have a seeded event with link to event homepage in the edit note and want to check something. Will spare some clicks.
monkey[m]
<mayhem[m]> "I think the right thing to do is..." <- mayhem: Halo! Question about the format of the service-status endpoint. For alerts stuff, is it better to have a playlists_age dict with each patch as a key in there, or to have everything in a flat dict?
[listenbrainz-server] 14MonkeyDo opened pull request #3074 (03master…playlists-service-status): Service status: monitor recommendation playlists separately https://github.com/metabrainz/listenbrainz-serv...
adhawkins has quit
adhawkins joined the channel
mayhem[m]
<monkey[m]> "mayhem: Halo! Question about the..." <- > <@monkey:chatbrainz.org> mayhem: Halo! Question about the format of the service-status endpoint. For alerts stuff, is it better to have a playlists_age dict with each patch as a key in there, or to have everything in a flat dict?
[listenbrainz-server] 14MonkeyDo opened pull request #3075 (03master…artist-search-disambiguation): LB-1707: Add sort name and disambiguation in artist search pages https://github.com/metabrainz/listenbrainz-serv...
zas[m]
Yes, but it has at least one issue: 2023 -> it will create one entry per year, we expect to use tags for that. Imho we should change the approach
monkey[m]
I think we only want to monitor some of those, and not the year in music ones, so these will probably be removed
zas[m]
if we can drop the per-year thing it would work
monkey[m]
OK, thanks for the confirmation
is dash vs. underscore in the playlist names an issue? might make sense to format like the rest?
zas[m]
nope, that's normal, because you have an object named playlists_age, and it creates fields for each property in it. but metrics are flattened. Another approach (likely beter) would be to have one measurement named lb_playlists_age, and use a tag for each playlist. It would help if the number of playlists increase (or if their name change)
Could it be provided on another endpoint?
like /playlists-status
so we could have a tag playlist=top-discoveries-for-year for example and a field age for all, it would also ease adding more metrics about playlists
jasje[m]
monkey: tried the app with background activity set as unrestricted?
salaxceitor[m] has quit
monkey[m]
jasje: I'll try now 😊
pite joined the channel
the4oo4 has quit
the4oo4 joined the channel
reosarevok[m]
yvanzo: there's a "# HELP process_cpu_seconds_total Total user and system CPU time spent in seconds."
I guess that's something that would be useful
There's also this, but not sure how useful they are:
reosarevok[m] sent a code block: https://matrix.chatbrainz.org/_matrix/media/v3/download/chatbrainz.org/oThTIEeIYpluGGisqMqiKkIB
That seems to be about go not psql though :/
Yeah, I think all that is about prometheus itself, heh
I wouldn't be surprised if zas already does all of this elsewhere though :) zas, do you already have a good way to monitor psql's cpu and memory usage and can we steal it :p
Basically we want to have good ways to get stats on our containers for mb-docker, particularly in the short term to test the indexer (but it'd probably be great to have a way to monitor it in general anyway)
Yeah, but I have no idea where docker_container_cpu comes from for example - there's a ton of extra jobs in our prometheus config
Well, it seems to come from influxdb but I assume it gets there in a different way...
Or I guess maybe it's unrelated to prometheus entirely and it comes through influxdb instead?
It feels like we should bother one of the sysadmins to set up this rather than ambling and trying to figure out how to do it tbh :)
yvanzo[m]
They (he) are (is) overwhelmed already, it is definitely very useful to have their (his) help on this, but we can't just put everything on their shoulders.
zas: How does this data get stored in InfluxDB? Is there another way to get it directly through Prometheus?
reosarevok[m]
I mean, I get that, and I have nothing against learning on the job, it's just I'm not sure the amount of questions I'd need to keep asking would be less effort for them than just doing it and it will cost MeB a lot more... 😅
But we'll see
(Because I'll struggle for 10 times longer I mean, not because I charge more)
yvanzo[m]
reosarevok: There are other metrics which are not system related that you can look into instead.
I updated our testbed doc accordingly.
zas[m]
It all depends, we have both influxdb & prometheus db, and there's a plugin for telegraf to read prometheus format (https://github.com/influxdata/telegraf/tree/mas...) and it can send them to influxdb
But we also have telegraf postgres input plugin already
yvanzo[m]
zas: We are looking for the simplest setup. We don't need alerts and notifications, just collecting metrics for the time of a performance test.
If we can avoid setting up influxdb, that would be great.
Btw mayhem: I have seen your homepage wording email, I am just working on YIM atm 👍 thanks for that, it will light a fire to try do a homepage design over Xmas break
salaxceitor: the project logos were created in Adobe Illustrator by monkey
salaxceitor[m]
Aerozol (@_discord_537897767968833566:chatbrainz.org) thanks for the reply, have a good one
