Alexis Kinsella created CAMEL-5994:
--------------------------------------
Summary: CamelJob implementation rely on some inconsistent
endpoints LRU cache - Camel endpoints may not be called on quartz fire after
some days of uptime
Key: CAMEL-5994
URL: https://issues.apache.org/jira/browse/CAMEL-5994
Project: Camel
Issue Type: Bug
Components: camel-quartz
Affects Versions: 2.10.3, 2.9.5
Environment: Quartz is in deployed in cluster mode.
Reporter: Alexis Kinsella
Priority: Critical
CamelJobs rely on some endpoints LRU cache from camel context. If i'm not
wrong, since LRU Cache relies on SoftReferences, Garbage Collector can decide
to evict some entries and especially QuartzEndpoints.
It happens that sometimes endpoints evicted from LRU Caches are
QuartzEndpoints. Since QuartzEnpoints are searched from CamelJob in LRU cache,
they are not reachable anymore.
The side effect, is that CamelJob isntance initializes a new QuartzEndpoint,
but loadbalancer is not initialized since it is initialized from QuartzConsumer.
The result is that quartz triggers jobs that instanciate Quartz endpoints with
unintialized load balancers which are unable to call processor from the routes
associated.
I have joined a fix that rely on the search of QuartzEndpoints from initalized
routes of Camel context.
Fundamentally, the search of endpoints should not rely on endpoints LRU cache.
It is a broken design.
Hope it helps.
By the way, there are other issues related to QuartzEndpoint load balancer
initalization: If quartz starts before startup of Camel routes you will have
the same issue: Correctly initialized QuartzEndpoints are not in LRU cache
since Camel has not started routes (and so endpoints and consumers). Quartz
will trigger job, and camel job will search for QuartzEndpoint and put it in
LRUCache with uninitialized load balancer.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
