Retrying connection to AMQP broker about amqproxy

[carlhoerberg]

I know what the issue is, I'm working on a fix, i think it will be ready tmrw. https://github.com/cloudamqp/amqproxy/releases/tag/v0.2.3 does not have the same issue as the current 0.3.x line.

[carlhoerberg]

With which client was this? After continuous publishing? Doing something else on the connection? The more information the easier it will be to replicate, and then fix.

[s-maj]

Right, this is mostly Celery(4.2.1)/Kombu(4.2.1)/PyAMQP(2.3.2) and one Pika app (latest version) It looks like apps can connect but cannot send or receive any messages.

For the Pika:
Connection close detected; result=BlockingConnection__OnClosedArgs(connection=<SelectConnection CLOSED socket=None params=<URLParameters host=10.240.9.151 port=9999 virtual_host=platform-stress ssl=False>>, reason_code=505, reason_text='UNEXPECTED_FRAME - expected content header for class 60, got non content header frame instead')

For Celery/py-amqp I think it's about heartbeats that being send by default by producer (it can't be disabled atm):
Unrecoverable error: ConnectionForced(None, u'Too many heartbeats missed', None, u'')

but also are some ambiguous messages like:
CRITICAL/MainProcess] Unrecoverable error: ConnectionError(302, u'UPSTREAM_ERROR', (0, 0), u'')

[carlhoerberg]

Both the heartbeat and UNEXPECTED_FRAME should be fixed in v0.3.3: https://github.com/cloudamqp/amqproxy/releases/tag/v0.3.3