shashank created CAMEL-27083:
--------------------------------
Summary: camel-scheduler - a consumer that is stopped and started
again never polls, and its stop can shut down the thread pool of the other
routes of the same scheduler
Key: CAMEL-27083
URL: https://issues.apache.org/jira/browse/CAMEL-27083
Project: Camel
Issue Type: Bug
Components: camel-scheduler
Reporter: shashank
Assignee: shashank
h3. What happens
The consumers of a {{scheduler:name}} endpoint share a thread pool that the
component keeps per name with a reference count.
{{SchedulerEndpoint.onConsumerStart}} takes a reference only when the consumer
has no scheduler and no thread pool yet, but {{onConsumerStop}} always releases
one.
# A consumer that is stopped and started again in place (JMX
{{ManagedConsumer}} stop/start, or a route policy that stops and starts the
consumer, such as the leader election policies of camel-zookeeper-master or
camel-hazelcast: "run the scheduled route on the leader only") still has the
pool and the scheduler of its first start. Its stop released the last
reference, so the component shut the pool down. The second start takes no
reference and schedules the poll on the pool that is shut down: with the
default thread pool profile (CallerRuns) the task is dropped without an error.
The route reports Started and never polls again.
# With two routes on the same scheduler name, a stop and start of one route and
then a second stop releases the reference of the other route: the pool is shut
down under the other route, which stops polling too.
# A consumer that has its own thread pool ({{scheduledExecutorService}}) or its
own scheduler never takes a reference, but its stop releases one, which can
shut down the pool of the other routes of that name.
A route stop and start through the route controller creates a new consumer, so
it is not affected.
h3. How to reproduce
{{SchedulerConsumerRestartTest}} (camel-core): a route
{{from("scheduler://solo?initialDelay=0&delay=10")}}; after the first poll
{{ServiceHelper.stopService(consumer)}} and {{startService(consumer)}}: the
consumer's thread pool is shut down and no poll comes. Two more tests cover the
shared pool (stop/start/stop of route a, route b on the same name stops
polling) and a consumer with its own pool whose stop shuts down the pool of the
other route.
h3. Fix
The consumer remembers that it took a reference to the component's pool. Only
then its stop releases one, and its next start takes a new reference and gives
the new pool to the scheduler it created at its first start.
_Filed with Claude Code on behalf of allthingssecurity._
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)