[rabbitmq-discuss] [Malformed Packet] in wireshark after a Channel.open frame

[Martin Sustrik]

Hi Bruno,

When I wrote the wireshark dissector for AMQP there wan no AMQP/0-9-1 by 
then, so it dissects AMQP/0-9. Maybe something have changed between two 
version of AMQP that makes it believed the frame is malformed?

Martin

On 10/26/2010 05:41 PM, Bruno Thomas wrote:
> Hello rabbitmq list,
>
>
> We try hard to understand what's happening with our integration tests :
>
> When we don't send http request to our application after, say 10 to 15
> min, when the first request is occuring, (sending stats messages to
> rabbit), the shutdownListener is called with the cause
> "java.net.SocketException: Connection reset".
>
> The shutdownListener is launching a thread that is trying to reconnect
> with rabbit (when its cause is not initiated by the application), and
> during that time, the messages are put in a queue. When the connection
> is reestablished, the messages in the queue are sent. We did this
> reconnection thread because we needed to restart the webapps (messages
> producers) when the rabbitmq broker was shutdown.
>
> We did tcpdump traces, and when we analysed the frames, we saw that :
> - there were not TCP connection resets
> - whenever there was Channel.open frame, the next one is tagged as a
> malformed packet by wireshark (the Open-Ok)
>
> Is it normal to have these malformed packets ?
>
> Then the connection is reestablished and the messages are sent and
> received by consumers.
>
> The problem is that the 2 first basicPublish (before the
> shutdownListener is called) are sent without IOException so the
> messages are lost. The consumers receive null messages.
> Did you see this kind of situation ?
>
> Thank you for your answers.
>
> Bruno Thomas
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss