These are chat archives for fanout/pushpin

25th
Oct 2018
Justin Karneges
@jkarneges
Oct 25 2018 01:21
@yukw777 my config is the default (examples/config/pushpin.conf)
Justin Karneges
@jkarneges
Oct 25 2018 01:28
@krishnaarava the control port only accepts requests in UTF-8. however, if you want to publish UTF-16 data to clients this is certainly possible
cd-rnukala
@cd-rnukala
Oct 25 2018 01:36
for example for this one @jkarneges
Justin Karneges
@jkarneges
Oct 25 2018 01:36
@yukw777 hanging publish request on mac is surprising, but pushpin's network code is slightly different on mac so I'm not sure how seriously to take that issue. all our real world use and benchmarking has been on linux
cd-rnukala
@cd-rnukala
Oct 25 2018 01:36
for @krishnaarava quesiton

curl --header "Content-Type: application/json" --request POST --data '{

"items": [
{
"channel": "test",
"formats": {
"http-stream": {
"content": "汉语"
}
}
}
]
}' http://host5561/publish

we set content-bin as Base64
and where will the content go?
Justin Karneges
@jkarneges
Oct 25 2018 01:38
set content-bin to a string of Base64-encoded UTF-16 data
cd-rnukala
@cd-rnukala
Oct 25 2018 01:38
ok
understood thanks
will test it
Peter Yu
@yukw777
Oct 25 2018 15:01
@jkarneges in what way is networking for Mac different?
Justin Karneges
@jkarneges
Oct 25 2018 17:58
@yukw777 well nothing different in pushpin specifically, but it uses the qt library for networking which I believe uses epoll on linux and something else (kqueue?) on mac. there are also potential middle layers there, like on linux I believe qt normally wraps glib these days
Peter Yu
@yukw777
Oct 25 2018 18:40
@jkarneges just to give you a quick update, I ran everything (pushpin, my backend and loadtest) on an ec2 instance. Seems to be doing a lot better than my macbook pro. However, when it got to our beefy load test, it still struggled albeit a bit better than my macbook pro. When I set message_rate to 0 in pushpin.conf, the message loss rate basically halved. And now I'm trying to replicate this on kubernetes.
what does message_rate exactly do? The behavior we observed is that it buffers messages to be published under heavy load?