#metabrainz

brb

well, the general issue is that it's unclear what exactly people want

then ask them what they way.

*want

chirlu` bitmap I don’t see anything in http://blog.musicbrainz.org/?p=3474 that would be related to that

some dated release appear random order at the end of collection list, where all the undated release are

see above for examples

I created a ticket, but bitmap is the only one who responded

zas: I'm around, what did you want? :)

reosarevok: just test mb.org

Non-beta? ok

(gah, someone please fix that thing with "stop using beta site" throwing you in the front page)

i did few experimental changes as we discussed yesterday

basically 4 servers are handling ws requests, and one the website

bitmap: is it possible somehow to make it so the "new edit notes" banner won't appear in prod after dismissing it in beta and viceversa?

jesus2099: can you move back to non-beta for a bit too and see if it feels any faster?

You know better than me how it used to be :)

you’re not getting me back there !

 ;p

the mass merge recordings are blazing fast (on beta), I’ve never seen that.

We're trying to test if new changes to the set up made a difference or not :)

next time I have a release to merge recordings, I’ll try in main server to see… what has been changed in main server ?

"basically 4 servers are handling ws requests, and one the website"

So it has a dedicated server now, we want to see if that makes things better

Anyone would like to unassign me from the /edition/:id task. It was done like 2 days ago and I would like to someting other now :)

i even written a script so any mb links of any kind (http/https, main/beta) will always link to what i want

reosarevok: GRREAT ! !

stanislas: are you waiting for review there or?

chirlu`: on my 13th page I have some releases with dates in the middle of release without

chirlu`: wait… what ?

Yes, but not all of the release events have dates.

chirlu`: oh i see !

e.g. https://musicbrainz.org/release/d8f8d195-b0ce-4...

shouldn’t it take the earliest dated release event, then ? (as a change proposal)

reosarevok: Yep

By whom? LordSputnik, review stuff! ;) (if it's him)

chirlu`: do you think what i say is logical ? should i make a ticket ?

No idea at this point. :)

I mean, it probably makes sense that if they have at least one date they'll sort by that

reosarevok: Yeah, LordSputnik - he's done a great work yesterday (actually it was today partially) reviewing my recent task - I think that I could do something other today :)

zas: seems like submitting edits is a bit slow

(but that might be a DB server thing too? Dunno)

Yes, submitting is more dependent on DB than anything

The gain is mostly when browsing

reosarevok: so ?

For me it seems much better, but i'll confirm with measures once we'll have the right setup

stanislas: Is this something that you've done in the past? (unassign then reassign)

(I'm a bit worried about what happens if someone else claims the task)

chirlu`: pingu is serving website alone atm, its cpu usage is quite high, ruaok thinks it is mostly due to template engine, do you know how we could improve that ?

reosarevok: These are the bookbrainz-ws testing tasks

Just give me the link :)

that was by far the smallest, but useful, ferreteria I've ever been to.

reosarevok: I've done 8/10 of these tasks and I've unassigned many times

I guess you know what you're doing then :) Done!

reosarevok: #YOLO

reosarevok: thanks :)

ruaok: as expected the gzip thing have no impact

s/have/has/

reverting

thanks reosarevok, i have to remember writing a ticket tomorow

reosarevok, have you had a chance to watch the tutorial?

I guess he didn't like that tab very much.

gcilou: on it :)

great, thanks!

One small thing is that annotations are not required (and often not needed) but the way you say it, it sounds like we would like people to write stuff there for every release :)

(even though you don't actually fill it, which should be a good hint, but still)

Ok.. I could add a pop up message that's like "Hey, you don't have to fill all these in"

Ok, kind of an actual problem now :) You say to enter Wikipedia sites on the links tab. Those actually go on the release group, and can't be added from here. While we really should allow doing that, and just storing them at the right level (and there's a ticket for it) right now if they try, they'll get an error message (I think) :)

OK..so it'll let you put them in but not publish it? (I did test putting one in..)

i think..

Hmm.

Other than that the release information section seems pretty good :)

Hmm, ok - so, it does let you paste it, it doesn't error as such, but there's no Wikipedia option it seems

I thought it just blocked it

Yeah, I thought I did label it for some reason..maybe the how-to has it in there

Dunno. Not a big issue, anyway - we can always add a note to the video if needed

gotcha

But we should clarify it one way or another so that people don't try :)

LordSputnik: ping

One thing to mention on "Manual entry" is that you don't *need* to enter an artist for every track - it will use the release artist by default in that case

Heh, I do like the ending too :)

Funny how you pasted one release link on the first step, but the final release has a different one :D

But that's the kind of nitpicky thing only an experienced MB editor should care about, heh

So all in all I think it's pretty good!

Let's see what Freso has to say if anything, but I'm happy with this as it is plus a couple notes over the video in youtube I think (since changing the actual spoken parts seems like more effort than it deservers)

deserves, even

Yeeah, I redid the first couple videos and not the last couple..I don't think anyone who need a tutorial will notice though..

Yeah, I think you're probably right :p

Gentlecat: I've claimed and submitted the "CB: Fix loading issue on Spotify mapping page when album info is not returned (MetaBrainz Foundation)" task.

yes, I receive notifications about it

Gentlecat: Good to know after 1 month of gci :) I thought that you receive maybe e-mails or something.

Gentlecat: I've changed one dictionary addition to use dict.update()

stanislas: we receive emails, but I suspect all of us check our emails all the time

(I know I do :) )

even though CPU usage on pingu is pretty high, the site response is much much better than before.

:)

Guys, does anyone know what's the priority on fixing http://tickets.musicbrainz.org/browse/MBS-8726 ?

Is it something that will get fixed eventually or is it on the ASAP pile?

I restart memcache about 10 minutes after I do an update, each hour

but that’s not a great solution

not on the radar just yet.

Yee, there are ways around but it weans that someone needs to keep an eye on it all the time. Which in our case is not ideal… :(

at least not that I know of. speak with chirlu` and bitmap and ask if it is on their radars

The reason I'm asking is that it's been flagged as quite a high priority for us so I'm trying to work out what's the best way to go about fixing it.

ok, will do. Thanks ruaok!

np

alastairp: shall we?

reosarevok: it's not easy since beta uses a different redis than prod, but...I think we could combine them eventually

or does it

nm, they're already on the same server with a different prefix. so easier than I thought

UmkaDK: it's on my radar, since we need it too. would a flag that clears memcached after each packet work for you?

ruaok: so what should we do with the blog post? I can outline some parts if you want

Hey bitmap

hey

i did few experiments you need to be aware of ;)

atm pingu is serving website, and others just ws

I saw mention of them. and the site seems snappy :)

yes, but there are few things we need to decide

first, we can't run on one server only...

too much load?

yes, but the main issue is the failure

i think we can work to reduce cpu load

check stats about pingu and web servers status

to have an idea

i just change the number of workers on pingu, setting 50 for website and 10 for ws (the reverse of others), so no service was stopped

i didnt save any config yet

musicbrainz-full https://www.irccloud.com/pastebin/cDRvImPq/

musicbrainz-full-include https://www.irccloud.com/pastebin/Y0OIoJCt/

this does the trick

pingu is serving 7% of requests

but output more data than any other server

and its cpu usage is a bit high

so the question is how can we optimize this

would it make sense to have one of the machines in both clusters?