Handling occasional timeouts or disruptions about tinder HOT 8 CLOSED

Have you looked into why Twitter::JSONStream isn't reconnecting for us? I believe it should be.

from tinder.

By default Twitter::JSONStream has no special behavior around a response with a non-200 status code, but it allows you to specify an on_error callback to handle this case. This commit introduces an error handler which raises an exception in these cases. Removing the raising of the exception may be necessary but also may not be sufficient to allow Twitter::JSONStream to reconnect in those cases. Perhaps @joshwand who wrote that commit may have some further insight.

from tinder.

@noahd1: I think you're on the right track that removing the raise of the exception might allow Twitter::JSONStream to reconnect. Feel free to play around and submit a pull request.

from tinder.

As I suspected, it is necessary to remove that exception, but not entirely sufficient, (at least in cases that I'm seeing). Unfortunately, this is rather difficult to test, as the problems seem to arise after long periods of inactivity. I do see a couple of reconnection attempts, and then I get this stacktrace:

2011-10-09 23:56:43 | FATAL  | CampfireBot::Bot | Unhandled exception while listening to room devops: Could not parse data entirely
/var/lib/jenkins/campfire-bot/vendor/ruby/1.8/gems/twitter-stream-0.1.14/lib/twitter/json_stream.rb:121:in `<<'
/var/lib/jenkins/campfire-bot/vendor/ruby/1.8/gems/twitter-stream-0.1.14/lib/twitter/json_stream.rb:121:in `receive_data'
/var/lib/jenkins/campfire-bot/vendor/ruby/1.8/gems/eventmachine-0.12.10/lib/eventmachine.rb:256:in `run_machine'
/var/lib/jenkins/campfire-bot/vendor/ruby/1.8/gems/eventmachine-0.12.10/lib/eventmachine.rb:256:in `run'
/var/lib/jenkins/campfire-bot/vendor/ruby/1.8/bundler/gems/tinder-e3c8012dc127/lib/tinder/room.rb:151:in `listen'

The error causes the program to exit of course. Any ideas on what this might be?

from tinder.

I'm having the exactly same problem. I can reproduce this 90% of time just by sleeping a certain amount of time (eg. 10 seconds) in my listen callback.

from tinder.

Yes, I can concur with @massive that some kind of timeout causes the problem for me, too. I found that the effective delay is because I'm using a blocking network connection talking to another service to feed the campfire stream. I switched to using Thread to make it non blocking and it's working fine now, if that helps.

from tinder.

My remedy for this was to execute code via EventMachine. Eg. executing long running shell commands with EM.system fixed the issue.

from tinder.

Hi there! We're doing some open source maintenance work and closing old issues. If this issue is still applicable or you would still like it to be addressed, please re-open the issue by April 17, 2017. Thanks!

from tinder.