poll timeout increase for near realtime outgoing delivery

[youngj]

poll timeout increase for near realtime outgoing delivery

[youngj]

My guess is that this would already work. EnvayaSMS uses a connection pool that allows two concurrent HTTP requests. You could set the EnvayaSMS poll interval to 1 minute and then configure your server to hold the request for ~55 seconds. Does that work?

[youngj]

Oh, I see, that probably doesn't work now because the HTTP socket timeout is set to 10 seconds. I suppose we could just increase the timeout to 60 seconds.

[magixbox]

Yes, the timeout is too short at the moment. Verified.

[magixbox]

I can see some potential issue with such a short timeout hardcoded in EnvayaSMS, but I do understand the purpose of having such short timeout: the shortest Poll Interval is 15 seconds, so the active poll process must be ended before another kick in.

The quick "fix" is to have timeout near the Poll Interval value. Say 10 or 20 seconds before the next process start.
So for us, we can set Poll Interval to 1 minutes, hold the server CGI for 40 seconds, and get out clean.

The issue here is that, once EnvayaSMS got the SMS delivered on the first query, it will still wait for next schedule poll - which can be very long from the first one if the poll timer isn't based on "time between each poll".

[youngj]

Try installing EnvayaSMS-long-poll.apk on https://github.com/youngj/EnvayaSMS/downloads ... it has a socket timeout of 60 seconds, only runs one poller task at a time, and adds an option for a poll interval of 5 seconds.

So just set Poll Interval to 5 seconds, and hold the server CGI for < 60 seconds... let me know if it works for you.

[magixbox]

Yes it works. I've set to hold the CGI for 10 seconds, poll interval of 5 seconds, the app query the server and stay on for 10 seconds till the server release, and hit back the server immediately with subsequence query.

This is exactly the behavior that we were expecting. Sweet!

[youngj]

Great. Released as 2.0.4.