Bob Browning created CAMEL-7500:
-----------------------------------
Summary: Concurrent modification of exchange during retry leads to
futher processing of failed messages
Key: CAMEL-7500
URL: https://issues.apache.org/jira/browse/CAMEL-7500
Project: Camel
Issue Type: Bug
Components: camel-netty
Affects Versions: 2.13.1
Reporter: Bob Browning
When a exception occurs on a netty TCP channel such as ChanelClosedException
then there are two invocations of the producer callback.
If there is a redelivery handler configured this causes either two threads to
be added to the scheduled thread-pool which then compete or in the more common
case the first invocation adds the redelivery thread but in doing so clears the
exception from the exchange such that when the subsequent callback invocation
occurs it see's the event as a success and continues routing of the exchange.
Note this also seems to be a cause of negative inflight messages on the route.
The first callback invocation occurs in the ChannelFutureListener which is the
usual case.
The second callback invocation which comes from the ClientChannelHandler
registered in the DefaultClientPipelineFactory used by the NettyProducer.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
