metatech created CAMEL-8135:
-------------------------------
Summary: DeadLetterChannel moves message to DLQ before
FailOverLoadBalancer attempts the second node
Key: CAMEL-8135
URL: https://issues.apache.org/jira/browse/CAMEL-8135
Project: Camel
Issue Type: Bug
Components: camel-core
Affects Versions: 2.13.2
Environment: ServiceMix 5.3.0
Reporter: metatech
When a route is defined with both a "errorHandlerRef" and a failover load
balancer, the DeadLetterChannel considers the message as failed and moves it to
the Dead Letter Queue, before the failover load balancer could try the second
node.
In attachment, there is a a test driver that reproduces the problem. Two HTTP
backends are created : the first is down and the second is up.
A "timer" route calls :
1. a route with load balancing and *without* DLQ : this works fine. An HTTP
request is sent to the first node (which is down). Afterwards another HTTP
request is sent to the second node (which is up).
2. a route with load balancing and *with* DLQ : this does not work. An HTTP
request is sent to the first node (which is down). Afterwards, the request is
moved to the DLQ, without attempting the second node.
I suspect that the problem occurs because the error persists in the
"UnitOfWork".
The problem is reproduceable both in Camel 2.10.7 and 2.13.2.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
