in development, I don't think it's a problem that the api endpoint will be localhost/1/stuff
2017-05-29 14903, 2017
alastairp
since there should be nothing in the web part which starts with a number
2017-05-29 14959, 2017
iliekcomputers
Api compat is localhost/2.0/
2017-05-29 14918, 2017
alastairp
oh. hmm
2017-05-29 14929, 2017
alastairp
I don't think we thought that through very well...
2017-05-29 14949, 2017
ruaok
me personally, I am against this split.
2017-05-29 14958, 2017
ruaok
it makes everything much more complicated.
2017-05-29 14913, 2017
iliekcomputers
I kinda feel the same way
2017-05-29 14916, 2017
ruaok
and our gateways actually can split traffic based on URL, not only on domain.
2017-05-29 14916, 2017
alastairp
so, we originally thought about doing the split when kafka was pulling down the website
2017-05-29 14928, 2017
iliekcomputers
But there would be performance improvements?
2017-05-29 14935, 2017
ruaok
iliekcomputers: not really, not.
2017-05-29 14937, 2017
iliekcomputers
From the split
2017-05-29 14938, 2017
ruaok
no.
2017-05-29 14943, 2017
ruaok
zas: got a moment?
2017-05-29 14944, 2017
alastairp
and it was kind of a "oh, perhaps this would make it better"
2017-05-29 14901, 2017
alastairp
we didn't make any measurements to show that this would improve responsiveness of the site
2017-05-29 14900, 2017
alastairp
if we can balance to different servers / uwsgi instances from a gateway, I agree that perhaps it's not as necessary as we originally thought [at that time newhost didn't exist either, so LB was going to run standalone]
2017-05-29 14901, 2017
ruaok
it won't.
2017-05-29 14952, 2017
ruaok
I had planned for exactly this for MusicBrainz some 15 years ago.
2017-05-29 14911, 2017
ruaok
I put the API to mm.musicbrainz.org . And then ian and nikki made fun of me for years for doing this.
2017-05-29 14949, 2017
ruaok
personally, let's ditch that PR and update the docs and remove the API distinction.
2017-05-29 14946, 2017
ruaok
alastairp: I don't sign NDA, unless there is a gun to my head. I feel the same about pinky-swears.
2017-05-29 14950, 2017
ruaok
:-D
2017-05-29 14959, 2017
ruaok makes a bug
2017-05-29 14917, 2017
iliekcomputers
Not a problem with me if we don't get anything for the increased complexity of code, +1 for ditching the PR
2017-05-29 14934, 2017
alastairp
oh crap
2017-05-29 14937, 2017
_flow_ has quit
2017-05-29 14948, 2017
_flow_ joined the channel
2017-05-29 14954, 2017
alastairp
yeah, it's true. I asked you for the most serious of swears
2017-05-29 14911, 2017
ruaok
I made a bug instead. :)
2017-05-29 14924, 2017
alastairp
you know how easy it is to ignore a bug
2017-05-29 14930, 2017
alastairp
I should have insisted on the pinky swear
2017-05-29 14948, 2017
ruaok
I assigned to iliekcomputers, you're safe.
2017-05-29 14921, 2017
ruaok
anyways, silliness aside. Should we ditch the artificial seperation of flasks?
just so that you're aware of this distinction beween functions that only check data, and ones that create things
2017-05-29 14915, 2017
alastairp
perhaps in the future if you're doing something similar you can do it in a slightly different way
2017-05-29 14938, 2017
github joined the channel
2017-05-29 14938, 2017
github
[listenbrainz-server] mayhem closed pull request #174: LB-62: Begin work on splitting LB into two different flask apps (master...webserver/split-into-two) https://git.io/v9prH
2017-05-29 14938, 2017
github has left the channel
2017-05-29 14938, 2017
iliekcomputers
alastairp: I see your point.
2017-05-29 14956, 2017
iliekcomputers
alastairp: I'll add the DEBUG flag for now
2017-05-29 14937, 2017
_flow_ has quit
2017-05-29 14945, 2017
_flow_ joined the channel
2017-05-29 14928, 2017
alastairp
iliekcomputers: confirm with ruaok first that he runs his stress tests with DEBUG=True
2017-05-29 14930, 2017
_flow_ has quit
2017-05-29 14938, 2017
_flow_ joined the channel
2017-05-29 14904, 2017
ruaok
you're going to make iliekcomputers do a pinky swear, aren't you?
2017-05-29 14923, 2017
github joined the channel
2017-05-29 14923, 2017
github
[listenbrainz-server] mayhem closed pull request #180: LB-156: Reconnect to RabbitMQ if connection gets closed (master...webserver/rabbitmq-connection-reestablish) https://git.io/vHnaY
2017-05-29 14923, 2017
github has left the channel
2017-05-29 14958, 2017
alastairp
nah, I only do it to people who I don't trust
2017-05-29 14913, 2017
ruaok
smrt.
2017-05-29 14919, 2017
ruaok starts plotting revenge
2017-05-29 14956, 2017
iliekcomputers
ruaok: before you go all Count of Monte Cristo, we have DEBUG = True during the stress tests, right? :P
2017-05-29 14913, 2017
drsaunders joined the channel
2017-05-29 14913, 2017
ruaok is doomed
2017-05-29 14934, 2017
ruaok
now I am pretty much obligated to fuck this one up. :)
2017-05-29 14922, 2017
iliekcomputers
lol
2017-05-29 14940, 2017
jesus2099 joined the channel
2017-05-29 14942, 2017
github joined the channel
2017-05-29 14942, 2017
github
[listenbrainz-server] paramsingh opened pull request #182: LB-161: Add DEBUG guard when creating new accounts (influx-test-scribbles...pinky-swears) https://git.io/vHCPQ
2017-05-29 14942, 2017
github has left the channel
2017-05-29 14919, 2017
ruaok
alastairp: do you want to do the honors and merge this?
2017-05-29 14921, 2017
zas
ruaok: i'm there
2017-05-29 14932, 2017
ruaok
oh hai.
2017-05-29 14940, 2017
ruaok
we already decided stuff. wanna know what we decided? :)
2017-05-29 14920, 2017
zas
nope ;)
2017-05-29 14932, 2017
ruaok
smrt.
2017-05-29 14937, 2017
iliekcomputers
lol
2017-05-29 14946, 2017
iliekcomputers
Good work team!
2017-05-29 14912, 2017
ruaok
we're speshul.
2017-05-29 14942, 2017
ruaok
zas: FYI... the question was about putting the LB API under a different sub-domain from the regular web pages.
previously it was thought that this split would be beneficial so that we could split traffic to different servers and to isolate servers having problems (e.g. kafka)
2017-05-29 14957, 2017
zas
yes, that's my preference
2017-05-29 14959, 2017
ruaok
we asked what the benefits were for the extra complexity of having sub-domains.
2017-05-29 14916, 2017
ruaok
preference to have a sub-domain?
2017-05-29 14923, 2017
zas
yes
2017-05-29 14933, 2017
ruaok
well, we dislike it.
2017-05-29 14935, 2017
ruaok
lol.
2017-05-29 14938, 2017
ruaok
!m team
2017-05-29 14938, 2017
BrainzBot
You're doing good work, team!
2017-05-29 14943, 2017
zas
ahah :)
2017-05-29 14903, 2017
ruaok
it presents a lot of extra complexity to make this happen.
2017-05-29 14907, 2017
zas
ok let's list plus/minus then... i don't have a strong opinion
2017-05-29 14915, 2017
ruaok
please.
2017-05-29 14925, 2017
zas
how do you dns RR with a subpath ?
2017-05-29 14955, 2017
ruaok
how do our gateways do it now?
2017-05-29 14909, 2017
ruaok
my thought was that the gateways already know how to do this, so we don't have to add extra complexity.
2017-05-29 14949, 2017
zas
gateways aren't dns RR, but IP fallback, with complex switch over, and still a botlle neck, not really as scalable imho
2017-05-29 14914, 2017
zas
but let's say that's ok
2017-05-29 14927, 2017
zas
what is the extra complexity you talk about ?
2017-05-29 14915, 2017
ruaok
extra containers, extra uwsgi procs.
2017-05-29 14916, 2017
hibiscuskazeneko has quit
2017-05-29 14930, 2017
zas
from gateways pov, it is much simpler to configure a full domain (if we don't use a specific IP), and it is easier to filter/log/etc...
2017-05-29 14935, 2017
ruaok
easier to have everything in one container, since that maps cleanly do a dev env.
