#metabrainz

:sad beep:

R2 doesn't go where we need it to go.

CatQuest: r2 goes to the airport

And is the one I travel on the most

it ends with "deetoo"

"d2"?

Also, when I travel on the others, I start closer to the edge of the city so never pass through that station

ha

yeah, i suppose if someone wants to use the train for the airport T2 they can use the R2 and change at sants. also not bad.

gcilou: gets a point!

woohoo

Anyway. Too much talk about commuter trains

(I honestly thought the ":sad beep:" would make people go "ooooh")

alastairp: these are not the trains you're looking for

the train knowledge is strong in that one

hi there

sorry, we're not home.

i mean, I there. :)

ey

ruaok: you where?

I'm deploying a musicbrainz mirror

(i's also a pun)

yes i am ,a phun

so phunny

now i am a bed. so bedding

and I wanted to ask a question about the rate limit

very duvet

shoot, Freso

is it possible to tune or disable the rate limit?

I'm catcat! you're getting scnadinavians confused

our software as shipped does not have a rate limit feature.

I'm only planning to use the webservice V2

that is part of our setup.

so, no need to disable anything

oh allright

then I need to go and check again

but I'm getting a lot of 503 errors :(

now, did you setup the search server?

no just the mirror

that is your problem.

db replication

etc

oh thanks!

when you make a search, it uses our search servers.

which are rate limited.

will go now and check how to deploy the search on the same server

thank you very much!

np

btw

i installed redis and memcached on the same VM

do I need to enable it in the configuration file?

to make use of it?

possibly. have a look, the file is pretty self explanatory.

if you used default setups, it might just work.

ok

thanks :)

is it a good idea to run the server and the search server on the same VM?

depends on how much ram you have in your VM and how many requests/s you want.

we have 15 or so machines to handle 500 requests/sec

so, nearly everything is duplicated one way or another.

thanks for tip

?

ruaok: http://chatlogs.metabrainz.org/brainzbot/metabr...

i just started with the minimum (2 vcpus 2GB of RAM) and build on top of that

you'll want more ram.

with 2GB things will be very very slow

i was expecting that ;)

but to be honest we don't have a clue about how many reqs/sec we expect

so we will need to build a baseline

who is your target audience?

i'm participating in a OSS project http://www.emby.media

it's a media server like plex

fc7_: You should sign up on https://metabrainz.org/ so we can link you. :)

:)

I will discuss that with the project leaders, I'm just doing a small part

so we started using musicbrainz to get metadata for users music libraries

but due to the rate-limitation on the service, library scans for users takes a long long time

so the idea is to start with one server to create a baseline of what kind of investment will be to support all our users on our own mirrors

in terms of cloud compute services....

how many VMs, with which amount of resources....etc

Setting up mirrors is usually our recommendation too, FWIW.

absolutelly

now we need to see how many we need and if we can afford that

to keep up with the load at a decent service level

btw is there any documentation I can check regarding performance tuning?

like memcached and redis configurations

etc

?

i mean specific for musicbrainz server

zas, bitmap, chirlu, ruaok: ^

fc7_: We're struggling with performance tuning ourselves[1], so not really sure. :) ([1]: https://blog.musicbrainz.org/2016/05/16/state-o... )

not aware of any specific config tweaks we've made from the defaults for those

redis isn't performance critical, mostly just used for logins

just make sure DEVELOPMENT_SERVER is disabled in DBDefs.pm

done

same for CATALYST_DEBUG

and DB_STAGING_SERVER

anyway I think we will only be publishing the V2 webservice, so no logins or any other thing but queries for the webservice

to retrieve metadata

what are your use cases? who are you setting this up for?

memcached and redis configs matter less than RAM overall.

is for Emby users that will retrieve metadata to feed their music libraries

using the webservice

heh.

problem is I don't know how many users we are talking about.... (yes it's very early stages)

well, I guarantee you that 2GB won't make for an effective mirror.

hahaha

got that ruaok

srsly, the kodi realm of things tends to generate a LOT of hits.

how much memory do you suggest?

to start with

so many that we have to throttle them. we can't afford the resources.

8GB?

and I have a sneaky feeling that you're setting up a mirror because people are unhappy with our level of service currently.

nah they don't

and if that hunch is anywhere near right, then its not matter of how many GBs, but how many servers.

sorry, some of them are unhappy

let's do some math.

but they don't know about the rate-limit on mb servers....

if you have 10 users, each making 1 request a second (like our limit) that is 10 requests a second.

a VM that is running app server, database and search server isn't likely going to serve more than 10 requests a second.

really, I'm just guessing at shit to try and paint some sort of picture to help you along.

no worries man

I don't expect help on that

we can figure it out ourselves...

so, if you're setting up a mirror for 10 people, why not just use our service. (it recently got a lot better, btw)

when we start to see the servers melt down...believe me...

I'm trying to save you a pile of trouble, is what I am saying.

yeah I get it and I appreciate it

:)

if you come out and say: Look, we have a mirror and 1000 people come and try and use it, you'll have 1000 unhappy people.

so, try and ask more questions about expected use and then come back and we'll chat more.

zas would be another good person to chat to about this.

thanks man

another question (I was reading "the state of the onion" link)

this is 1 part of our hosting setup.

the machines that run the web front ends.

you guys have offices in Spain?

yep.

as soon as we get power from endesa we'll actually be in business. :)

so far I can power the fiber connection off the solar panel for a few hours a day. :)

lol