[
https://issues.apache.org/jira/browse/CAMEL-16927?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Piotr Klimczak updated CAMEL-16927:
-----------------------------------
Description:
My scenario:
# Http client -> Transactional servlet consumer
# Servlet route calls java bean
# Java bean calls another route using Camel Proxy and direct-vm
# CamelInvocationHandler calls direct-vm route and then hangs as route is
never executed, waiting for worker
At this point it is all deadlocked and it never ends.
From what I can see the problem happens in DefaultReactiveExecutor.Worker.
>From debugging I can see there are two DefaultReactiveExecutor.Workers at the
>point of deadlock.
At point 3, Worker number 2 executes SimpleTask which waits for it to finish-
this task is to execute bean as per step 3.
Bean execution then attempts to trigger another route using Camel Proxy and
direct-vm and schedules it on Worker number 2.
However immediately after scheduling on Worker number 2 the direct-vm
processor, Worker 1 executes executeFromQueue (triggered by countdownLatch
waiting for direct-vm call to finish) which then causes deadlock because task
added to Worker number 2 is never executed.
I have confirmed above in debugging, will try to come up with simple test case
as our scenario is rather complex and involves multiple projects.
Unfortunately for us this is blocker as our solution which was working
perfectly fine with camel 2 is not working.
was:
My scenario:
# Http client -> Transactional servlet consumer
# Servlet route calls java bean
# Java bean calls another route using Camel Proxy and direct-vm
# CamelInvocationHandler calls direct-vm route and then hangs as route is
never executed, waiting for worker
At this point it is all deadlocked and it never ends.
From what I can see the problem happens in DefaultReactiveExecutor.Worker.
>From debugging I can see there are two DefaultReactiveExecutor.Workers at the
>point of deadlock.
At point 3, Worker number 2 executes SimpleTask which waits for it to finish-
this task is to execute bean as per step 3.
Bean execution then attempts to trigger another route using Camel Proxy and
direct-vm, but because it schedules direct-vm processor to the same Worker 2
which is currently processing current task it all deadlocks.
I have confirmed above in debugging, will try to come up with simple test case
as our scenario is rather complex and involves multiple projects.
Immediately after scheduling on Worker number 2 the direct-vm processor,
Worker 1 executes executeFromQueue (triggered by countdownLatch waiting for
direct-vm call to finish) which makes me thinking that direct-vm call has been
added to the wrong Worker (sits in wrong queue).
Unfortunately for us this is blocker as our solution which was working
perfectly fine with camel 2 is not working.
> Route deadlock in DefaultReactiveExecutor/Worker in transactional route which
> calls Camel Proxy bean
> ----------------------------------------------------------------------------------------------------
>
> Key: CAMEL-16927
> URL: https://issues.apache.org/jira/browse/CAMEL-16927
> Project: Camel
> Issue Type: Bug
> Components: camel-core-engine
> Affects Versions: 3.7.5, 3.7.2
> Reporter: Piotr Klimczak
> Priority: Critical
>
> My scenario:
> # Http client -> Transactional servlet consumer
> # Servlet route calls java bean
> # Java bean calls another route using Camel Proxy and direct-vm
> # CamelInvocationHandler calls direct-vm route and then hangs as route is
> never executed, waiting for worker
> At this point it is all deadlocked and it never ends.
> From what I can see the problem happens in DefaultReactiveExecutor.Worker.
> From debugging I can see there are two DefaultReactiveExecutor.Workers at the
> point of deadlock.
> At point 3, Worker number 2 executes SimpleTask which waits for it to finish-
> this task is to execute bean as per step 3.
> Bean execution then attempts to trigger another route using Camel Proxy and
> direct-vm and schedules it on Worker number 2.
> However immediately after scheduling on Worker number 2 the direct-vm
> processor, Worker 1 executes executeFromQueue (triggered by countdownLatch
> waiting for direct-vm call to finish) which then causes deadlock because task
> added to Worker number 2 is never executed.
> I have confirmed above in debugging, will try to come up with simple test
> case as our scenario is rather complex and involves multiple projects.
> Unfortunately for us this is blocker as our solution which was working
> perfectly fine with camel 2 is not working.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
