Rushan Shaymardanov

Thank you for reply and for hyper ) > So there are two possibilities. One is that, while you might only allow 50 greenlets, the send greenlets might be starving...

`TooManyStreamsError` was not reproduced yet, but I have some debug logs. There are a lot of messages like this: ``` 2016-08-03 13:05:00,770:DEBUG:hyper.http20.connection:recv for stream 0 with set([1, ]) already present...

We enabled logging of streams count and currently maximum what we have is 12 streams: ``` 2016-08-04 13:39:43,959:DEBUG:backend.push.push_apns2:Have 12 opened streams ``` It seems strange because it rather far from...

We decided to decrease grenlets number to 10. Currently we have following in logs: ``` 2016-08-07 00:54:07,361:ERROR:celery.worker.job.on_timeout:Hard time limit (900s) exceeded for backend.tasks.async.send_push[378a60a9-54aa-460b-be66-79c875ff6239] 2016-08-07 00:54:07,647:ERROR:backend.push.push_apns2.apns_send_msg:Too many opened streams, 168 2016-08-07...

@Lukasa here it is https://www.dropbox.com/s/ly57q0vfq8r5jtt/%D0%9D%D0%B0%D1%87%D0%B0%D0%BB%D0%BE%20%D1%80%D0%B0%D0%B1%D0%BE%D1%82%D1%8B%20%D1%81%20Dropbox.pdf?dl=0 But seems there is no any messages from hyper except for one I mentioned above

@Lukasa sorry, here is right one https://www.dropbox.com/s/4ibew32vllgqmsv/lastlog.tar.gz?dl=0

BTW, in celery task we have something like this: ``` stream_id = connection.send(token, pl, topic=topic) status, apns_id, body = connection.feedback(stream_id) ``` So seems like execution stucks on getting feedback. Do...

@Lukasa currently, no. I'm trying to figure out how to get it and will write here if I have any success with it

Unfortunately, I still did not fugure out, how to get stack trace for `Hard time limit (900s) exceeded` error, so I made a trace dump of all greenlets in moment...

@Lukasa do you have any suggestions on how we can deal with it?