The IA one? IA’s API is not "CAA API". And in a MeB context, certainly not "the" API. That said, I’d be interested in seeing the code you used.
CatQuest
hi Freso! `😺`
🐾😿
zas
solr cloud is still unstable :(
it works somehow, but metrics show it isn't working as usual, response times are very bad, and each node exhibits different figures (load/disk activity/etc...), also there was a failure during 1 hour, but it seems to have partially recovered by itself...
ruaok
the failover is still disabled?
zas
nope, it's up, and working, though i switched to lb4 as master, and it seems to underperform compared to lb3, i may switch back again.
ruaok
but it didn't start randomly switching over during the last failure?
zas
the main issue was unsynced solr nodes, I just restarted them one by one, with proper delays and order
ruaok
what should I next steps be?
*oour
zas
the random switching didn't happen, and the floating IP shouldn't float too much now
ruaok goes to make another coffee
well, first, ensure the cloud is stable again
ruaok
so, the failover was a symptom of the instability?
zas
yes
solr cloud stability is very tricky, in this regard one additional node might help
if you don't restart them in the correct order and/or too fast, the whole thing enters in unstable state, keep trying to recover, sometimes succeeds, sometimes fails
ruaok
why do you think it is so unstable?
reosarevok
^ that sounds like Chernobyl again :p
zas
because numbers show it: not the same number of items in all replicas, not the same cpu/disk/network activity
btw, response times are awful too
I *think* it is recovering now, I manually rebooted each node, you have to always end with "leader" node (https://stats.metabrainz.org/d/T4MODrIiz/solr-c... shows which one it is), if possible and let enough time between each node restart for them to sync with others
in short, tricky stuff, that doesn't work as one could expect
plus it's not totally dynamic (adding and/or removing nodes isn't trivial), read zookeeper/solr cloud docs if interested in details
to ponder that, it works great most of the times
ruaok
and not running docker, right?
zas
nope, it's not really needed to add extra mess here ;)
ok, what can we do to automate the adding/removing of nodes?
because even if capacity is NOT the issue, debugging a system near capacity makes everything much more difficult.
zas
automate ? well... start to read docs about solr cloud & zookeeper, you may soon understand why it isn't *that* easy. I have all what is needed to do it, but you have to understand that adding a node requires configuration changes on other nodes.
Freso, they're quite similar, although unfortunately using your own IA keys musicbrainz doesn't like.
Freso, a reply to your notes, I'm talking about the relationship on the releases, will add a link to the rest
Mr_Monkey
akhilesh: I've been looking at good ways to document the API and serve those docs. I quite like the idea of using inline JSDoc in the code (like we do for the rest of the projects) and have automatically serve documentation pages based on that.
Fancy giving it a go? It's the right moment to start documenting as you go, to avoid having to pay it back later :p
D4RK-PH0ENiX has quit
akhilesh
Mr_Monkey: I agree, but my target is to complete lookup request with tests then document the code as well as api endpoint for lookup request properly, try to get it merge, then move forward to browse requests. Is it fine to you?
Mr_Monkey
Yes, that's fine. Just thought I'd post this setup before I forget :)
iliekcomputers
ruaok: the messybrainz test failure seems like a Jenkins thing, I'll check.
ruaok
Thx
iliekcomputers
also, monies arrived, thanks! :)
ruaok
yeyaha,
D4RK-PH0ENiX joined the channel
Cyna
bitmap: I checked in another browser and facing the same result
Freso
zarcade_droid: Rule of thumb: don’t require editors to look up an entity’s links themselves; present it to them directly in the edit note. Both to make voters’ lives easier (less effort to check the edit) but also to lessen "bit rot": the links listed for a given entity may change over time, but the edit note is (actually) permanent. An entity may also get deleted (intentionally or by accident) in which case it is no longer
possible to look up relationships, but the edits are still in the system, so someone may still come across the edit and want to double check it.
zarcade_droid: And the "API" for uploading to CAA is: https://musicbrainz.org/release/$MBID/add-cover... – anything else is unsupported. (I’d still like to see your code though. In general, it’s a good idea to disclose in edit notes when doing tool assisted edits, including a source to the tool.)
zarcade_droid
Freso, yeah, adding the link to the edits now but if the relationship is removed from the original then probably for a good reason and the one in the edit will also be useless, but shrug, if it helps to vote on the cover art I'm adding it
rsh7 joined the channel
Freso
It won’t be useless: it will tell what the source of the image was.
zarcade_droid
*was*
Freso
If that’s no longer considered a good source, then that might be an argument for removing the image. (Probably not on these cases.)
zarcade_droid
There's also the API endpoint "https://musicbrainz.org/ws/js/cover-art-upload/"
Freso
That’s an internal API that isn’t supported for others.
ws/2/* is the only supported API.
zarcade_droid
Given that the APIs aren't officially supported, I'm wary of providing any source
Freso, there's also the licensing, lack of comments and requirement to run scripts in a certain order with s***load of manual work to get the artists linked
so like, if you're interested in a certain component I can share that but it has no value for anyone else
rcombs
so, is the new search stuff ready to deploy yet, and is there any documentation on how to keep the indexes up-to-date?
ruaok
rcombs: yvanzo is the person to help you with that
rcombs
yvanzo: ohai
zarcade_droid
(just to separate questions nicely) a) What could be done to get bigfinish.com URLs normalized, type autodetected and displayed on the sidebar?
b) Release medium without tracks is probably soon doable - now comes the next step from it, would it be possible to do works related to a medium instead of a track?
