Details
-
Bug
-
Status: Open
-
Critical
-
Resolution: Unresolved
-
proton-c-0.24.0
-
None
-
None
-
Docker (Kubernetes) / Azure ServiceBus
Description
A python-instance based on v0.24.0 is used to receive and send messages from/to a Azure ServiceBus instance.
The affected software, which runs inside a Docker container and which is based on Python 3.6, is already used since a few weeks. From time to time the conatiner(s) may restart, which is not a problem.
The application receives exactly one message from the Azue ServiceBus and sends exactly one message back when the work is done. Qpid Proton is easy to use and always worked well, but this weekend I had a problem: The application still received messages and successfully handled them. It also sent and "answer"-message, but they never appeared in the Azure ServiceBus.
I started another process on one of the two available affected containers. This second process worked without any problems. As I found out a restart of the containers (more precisely: "affected" processes) solved the problem.
Unfortunately, it is very critical for the given application that sent messages are really received by the Azure ServiceBus. Or if this doesn't work, that it is possible to know that this doesn't work. Usually, if the connection is broken (or sending messages doesn't work) Qpid Proton handles this case (incl. logging).
Connection initialization:
self._connection = container.connect("amqps://a:b@xyz.servicebus.windows.net", shared_access_key_value="key", service_namespace="namespace" ), heartbeat=self.heartbeat_ms) def _get_session(self): session = self._connection.session() session.open() return session
Initialize receiver:
self._receiver = event.container.create_receiver(self._get_session(), self._source_queue_name) self._receiver.capacity = self.capacity
Initialize sender:
self._sender = event.container.create_sender(self._get_session(), self._target_queue_name)
Do I do something wrong? Or are there any bugs in combination with the ServiceBus known? Unfortunately, I was not able to collect much debug information during the runtime, but is there a possibility that I can collect useful information if this happens again?
It might be useful to know that during the period while this happend, the ServiceBus topology was not changed/updated.
Thank you very much