Thread lock on getKieSession()

[dino.antonelli]

Hi,

I need some support with JBPM 6. I'm using the 6.2.0.Final version in a multi-threaded environment. I exposed some JBPM process services as rest services. I configured the runtime environment to be PER_PROCESS_INSTANCE. I have a problem when I try to retrieve the kie session; sometimes (it happens a couple of time in a day with nearly 1000 users working...), when the application executes the following code:

			RuntimeEngine runtimeEngine = this.runtimeManager.getRuntimeEngine(ProcessInstanceIdContext.get(processInstanceId));
			KieSession ksession = runtimeEngine.getKieSession();

there's a lock that blocks the execution; below a stacktrace of the thread that remains locked:

service(HttpServletRequest, HttpServletResponse)

doPost(HttpServletRequest, HttpServletResponse)

processRequest(HttpServletRequest, HttpServletResponse)

doService(HttpServletRequest, HttpServletResponse)

doDispatch(HttpServletRequest, HttpServletResponse)

handle(HttpServletRequest, HttpServletResponse, Object)

handleInternal(HttpServletRequest, HttpServletResponse, HandlerMethod)

invokeHandleMethod(HttpServletRequest, HttpServletResponse, HandlerMethod)

invokeAndHandle(ServletWebRequest, ModelAndViewContainer, Object[])

invokeForRequest(NativeWebRequest, ModelAndViewContainer, Object[])

doInvoke(Object[])

invoke(Object, Object[])

invoke(Object, Object[])

invoke(Object, Object[])

completeTask(long, long, DocumentUpdate)

completeTaskByProcess(String, long, long, HashMap)

getKSession(Long)

getKieSession()

initKieSession(Context, InternalRuntimeManager, RuntimeEngine)

findKieSessionById(Long)

getConfiguration()

newKieSessionConfiguration(Properties, ClassLoader)

<init>(Properties, ClassLoader)

init(Properties, ClassLoader[])

<init>(String, ClassLoader)

<init>(String, ClassLoader, boolean)

getResources(String, ClassLoader)

getResources(String)

getResources(String)

getResources(String)

findResources(String, boolean)

getResources(String)

loadResourceLocal(String)

loadResourceLocal(String)

getResource(String)

doPrivileged(PrivilegedAction)

run()

run()

toURL()

getVirtualURL(VirtualFile)

getVirtualURI(VirtualFile)

getPathName(boolean)

isDirectory()

isDirectory(VirtualFile, VirtualFile)

doPrivileged(PrivilegedAction)

loadStatefulKnowledgeSession(Long, KieBase, KieSessionConfiguration, Environment)

loadKieSession(Long, KieBase, KieSessionConfiguration, Environment)

loadKieSession(Long, KieBase, KieSessionConfiguration, Environment)

buildCommandService(Long, KieBase, KieSessionConfiguration, Environment)

newInstance(Object[])

newInstance(Object[])

newInstance(Object[])

<init>(Long, KieBase, KieSessionConfiguration, Environment)

initExistingKnowledgeSession(Long, KieBase, KieSessionConfiguration, PersistenceContext)

loadSnapshot(byte[], KieSession)

unmarshall(InputStream, KieSessionConfiguration, Environment)

unmarshall(InputStream, KieSessionConfiguration, Environment)

readSession(MarshallerReaderContext, int, Environment, SessionConfiguration)

createAndInitializeSession(MarshallerReaderContext, int, Environment, SessionConfiguration, ProtobufMessages$KnowledgeSession)

createWorkingMemory(long, InternalKnowledgeBase, FactHandleFactory, InternalFactHandle, long, SessionConfiguration, InternalAgenda, Environment)

<init>(long, InternalKnowledgeBase, FactHandleFactory, InternalFactHandle, long, SessionConfiguration, InternalAgenda, Environment)

<init>(long, InternalKnowledgeBase, FactHandleFactory, boolean, long, SessionConfiguration, Environment, RuleRuntimeEventSupport, AgendaEventSupport, RuleEventListenerSupport, InternalAgenda)

createProcessRuntime()

newProcessRuntime(StatefulKnowledgeSessionImpl)

newProcessRuntime(InternalWorkingMemory)

newProcessRuntime(InternalWorkingMemory)

<init>(InternalWorkingMemory)

initProcessEventListeners()

addEventListener(KieBaseEventListener)

contains(Object)

containsKey(Object)

getEntry(Object)

It seems that the thread remain in wait in the following sychronized method of the AbstractEventSupport object in the contains call:

    public final synchronized void addEventListener(final E listener) {

        if (!this.listeners.contains(listener)) {

            this.listeners.add(listener);

        }

    }

Does anyone know anything about this issue???

Thanks

Dino

[dmarrazzo]

Hello Dino, race conditions are quite complex to debug. I suggest you to create a reproducer and open a case to the RH support.

[dino.antonelli]

Thanks a lot Donato. I know that the problem is complex, My hope was that someone else could have faced this problem before. Unfortunately the problem is coming out in production environment while everything is ok in the stress/performance environment. So I'm not able to reproduce it.

Dino

[dmarrazzo]

In this case, I suggest you to collect a set of thread dump (every 5 second), when the situation happens. https://access.redhat.com/solutions/18178 Then, open a case with this information. The support engineer will help you to find the root cause. An interesting test could be using this runtime strategy: PER_REQUEST

[dino.antonelli]

Hi Donato, I'll try the thread dump collection ...

Thanks a lot

Dino