19:35 PM
sjl
read and cry
19:35 PM
honza
i guess it's time write celery for humans
19:36 PM
kennethreitz style
19:36 PM
sjl
I think that's RQ
19:36 PM
ehazlett
sjl: yep
19:37 PM
honza
i still think amqp > python pickle
19:37 PM
sjl
pickle is a serialization format, amqp is a messaging format...
19:37 PM
ehazlett
honza: i thought celery still pickled to rabbit
19:37 PM
yeah
19:37 PM
sjl
yeah
19:37 PM
honza
what
19:37 PM
sjl
pickle is like JSON or EDN
19:37 PM
amqp is like the Redis protocol
19:37 PM
honza
hm
19:37 PM
sjl
or 0mq protocol
19:38 PM
you need one of each
19:38 PM
honza
right
19:38 PM
sjl
RQ uses pickle + redis
19:38 PM
celery uses pickle + (whatever protocol your backend uses)
19:39 PM
ehazlett
sjl: plus shit
19:39 PM
honza
i just feel like amqp is more of a standard and other tools play nicely with it --- but the same is probably true of the redis protocol these days
19:39 PM
ok, you convinced me
19:40 PM
sjl
honza: the original designer of AMQP has called it "fatally flawed"
19:40 PM
19:40 PM
er
19:40 PM
"fundamentally flawed and unfixable"
19:40 PM
same thing
19:40 PM
honza
welp
19:40 PM
sjl
heh
19:40 PM
honza
wtf
19:41 PM
ok, since we're talking... what about rabbitmq vs redis
19:41 PM
for pubsub
19:41 PM
ehazlett
i don't think rabbit does that
19:41 PM
sjl
19:41 PM
ehazlett: it can
19:41 PM
ehazlett
er wait yeah i think you can use fanout
19:41 PM
sjl
yeah
19:41 PM
ehazlett
sjl: yeah
19:42 PM
sjl
honza: redis is great until the single server you're running it on breaks
19:42 PM
honza
exactly
19:42 PM
sjl
honza: it's music easier to write something that works on one machine like redis than to write something that works as a cluster like rabbitmq
19:42 PM
honza
rabbitmq seems to be more resilient or something
19:42 PM
sjl
*much
19:42 PM
I can't fucking type
19:42 PM
honza
heh
19:42 PM
right
19:42 PM
sjl
rabbit is designed to work in a distributed fashion from the start, and it supports a lot more than the basic pub/sub stuff redis provides
19:43 PM
zeromq is another option, though much more mind-bendy
19:44 PM
honza
yeah
19:45 PM
i always thought zeromq was a nice abstraction and you could put whatever you wanted under it
19:45 PM
but i'm not even a real programme so what do i know
19:45 PM
sjl
honestly if Python had something like Clojure's (future) and used mostly immutable data structures I'd just use that for celery-like tasks
19:45 PM
honza
sjl: +10
19:45 PM
sjl
shoving tasks over the network is too much pain for too little benefit
19:46 PM
ehazlett has quit
19:46 PM
arthurdebert
all: back
19:47 PM
sjl
serialize, send over a wire, sit in a queue, deserialize -- in exchange you get... to maintain more things?
19:47 PM
just let the app servers do the work. if they start getting too busy, spin up more of them.
19:47 PM
but since Python threading is a joke that's not necessarily a good option
19:48 PM
janeted joined the channel
19:49 PM
honza
plus you have to have a copy of the entire repo on the worker node
19:49 PM
janeted
sjl - what is the current state of the staging branch?
19:49 PM
I'm only asking because the entire webdriver_testing directory seems to be MIA
19:50 PM
sjl
janeted: staging is now in the "before the big DMR merge" state
19:50 PM
19:50 PM
janeted
i did - just wasn't clear how that would affect things
19:51 PM
and so when you do the big merge, then all the changes will be back?
19:51 PM
sjl
janeted: basically all the main DMR code is on dev
19:51 PM
if you need to do anything related to the DMR it should go on dev
19:51 PM
janeted
ok - was just feeling a little scared
19:51 PM
sjl
and I'll handle making sure dev gets into staging at the appropriate time
19:51 PM
janeted
got it - thanks
19:52 PM
honza
19:52 PM
nicksergeant
aww sheeit
19:52 PM
sjl
nice
19:52 PM
we could ditch our CustomUser crap
19:52 PM
ehazlett joined the channel
19:53 PM
nicksergeant
yeah
19:53 PM
score
19:53 PM
sjl
or rather, plug it in properly instead of the hacked in way we do it now
19:53 PM
nicksergeant
you mean the hacked way in which every django app on earth does it?
19:53 PM
hehe
19:53 PM
specifically userprofiles was always a mess
19:53 PM
and felt very second-rate
19:54 PM
sjl
yep
19:54 PM
fernandotakai
that was a thing that i always complained
19:54 PM
sjl
oh neat, the verbatim template tag is in core now
19:54 PM
nicksergeant
wat? you want more than a name and an email for your users?! nevar!
19:55 PM
ooo
19:55 PM
19:55 PM
that's hot
19:56 PM
sjl
ehazlett: staging is still running pre-tern migrations, btw, so far it looks good
19:56 PM
ehazlett
sjl: awesome
19:56 PM
sjl
about 30% done with them
19:56 PM
I'm getting good at this process with all the practice, lol
19:56 PM
probably a good thing
19:56 PM
arthurdebert
all: anyone deploying to prod in the next hour or so?
19:57 PM
nicksergeant
arthurdebert: not
19:57 PM
sjl
arthurdebert: not me
19:57 PM
ehazlett
arthurdebert: not me
19:57 PM
nicksergeant
arthurdebert: i
19:57 PM
arthurdebert
tks folks
19:57 PM
nicksergeant
19:57 PM
niiiiice
19:57 PM
timezone conversions built-int
19:57 PM
nigelbot________
AST : 15:54:12 AST-0400
19:57 PM
Sao Paulo : 16:54:12 BRT-0300
19:57 PM
EST : 14:54:12 EST-0500
19:57 PM
UTC : 19:54:12 UTC+0000
19:57 PM
PST : 11:54:12 PST-0800
19:57 PM
nicksergeant
that's awesome
19:57 PM
lol
19:58 PM
honza
nicksergeant: that is awesome
19:59 PM
sjl: a subtitles.pipeline question for you: do you have any thoughts on how the visibility attribute should be inherited? right now it seems to ignore previous states
20:01 PM
sjl: ie when you add a new sv, it doesn't look at the parent, it just passes the attribute down the chain
20:02 PM
sjl: should that just be entirely on the caller of add_subtitles()?
20:02 PM
arthurdebert
honza won't that make things more opaque, as in the caller should know what is doing
20:02 PM
honza
i'm just thinking out loud
20:02 PM
arthurdebert
honza yeah me too,
20:03 PM
sjl
honza: yes, the caller is responsible for deciding what to do with visibility
20:03 PM
honza
k
20:10 PM
DarrenBridenbeck has quit
20:12 PM
heidiv joined the channel
20:12 PM
launchpad-bot
Task complete: demo:amara,prod-test proxy_user:arthur deploy
20:20 PM
arthurdebert
sjl is there a ballpark for when staging will be on DRM code? I'm just trying to coordinate with partners when they'll be able to do work on staging again
20:21 PM
sjl as in "later today" "tomorrow" or "later this week"
20:21 PM
sjl
arthurdebert: however long it takes the tern migrations to run, which I'll hopefully be able to start before I have to leave for my bass lesson. usually takes 3-4 days for those I think, so Monday is probably a safe bet?
20:22 PM
arthurdebert
sjl k, that sounds like a plan
20:38 PM
20:38 PM
honza
arthurdebert: seems suspicious
20:39 PM
arthurdebert
honza hm… udacity is reporting getting 500 errors on a video resource request, can't find anything on sentry
20:39 PM
honza
arthurdebert: it's probably on the root logger
20:42 PM
arthurdebert
honza can't find it there neither
20:43 PM
honza
arthurdebert: hmmm
20:43 PM
arthurdebert: is it happening right now? sometimes those messages are late coming in
20:44 PM
arthurdebert
honza it happend a couple of hours ago
20:45 PM
honza
arthurdebert: getting timeouts on sentry, ugh
20:45 PM
arthurdebert
honza arg
20:46 PM
ehazlett
honza: yeah large queries result in timeouts
20:46 PM
honza
software is hard
20:46 PM
arthurdebert
honza software is too brittle
20:47 PM
ehazlett
honza: wow that sounded mean
20:47 PM
honza: sorry -- i didn't mean it that way
20:47 PM
:)
20:47 PM
michelled_ joined the channel
20:47 PM
honza: i just meant sentry is a pita sometimes
20:47 PM
honza
ehazlett: huh? i don't get it :()
20:47 PM
:)
20:47 PM
ehazlett
honza: ah sorry - i re-read it and thought it sounded mean :)
20:48 PM
honza: disregard then