Closed adamvduke closed 8 years ago
Hi @adamvduke, we can totally make it smaller & configurable. Give me a couple of days and I'll issue a change. Would you mind giving me your feedback then?
Also, a side question: how many pushes are you doing / day?
The current throughput for this particular service is ~10k/day, but the expected throughput is 10's of millions. Happy to give feedback on any changes or help if needed.
This commit gets rid of any delay, and will instantly unlock the blocking call ASA the response is received.
Can you try master NetHttp2 with Apnotic and see if this works for you? I can build a temp branch of Apnotic that uses master NetHttp2 if that's a hassle for you.
Hi @ostinelli,
The temp branch of apnotic would be helpful because the version of net-http2 has changed from ~> 0.9.2
in source to ~> 0.10.0
which causes some issues with bundler when trying to resolve correct versions based on apnotic's current gemspec.
cc @adamvduke
@acceber95, @adamvduke
I realized that my proposal is undoable, since I cannot specify a dependency to a gem's master branch.
Just set in your Gemfile
:
gem 'apnotic', '~> 0.9'
gem 'net-http2', git: 'https://github.com/ostinelli/net-http2.git', ref: 'a86074ce406db68190ec50a344d521d71ec91ee6'
NetHttp2 0.11.1
and Apnotic 0.9.2
have been released.
I haven't heard back from you @adamvduke, let me know if this solves your issue.
I will try to test it out this week.
Hello @adamvduke, any news on this?
We've seen the timing drop down to around 75ms on average and it is not a hard floor anymore. We haven't seen a noticeable change in response codes or exception rates from Apple, which seems to indicate things are working well.
Thank you for the feedback, @adamvduke. If you end up using Apnotic for the millions pushes that you were referring to, please let me know :)
Hi, first I wanted to say thanks for contributing the gem! We're using apnotic to try out the HTTP2 APNS endpoints and see the time to complete a request has a consistent floor very near 100ms.
I was wondering how the times to sleep were decided on here and here. Is there a specific reason 100ms is used and/or can it be made smaller or configurable?