[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ecf-dev] OOM Error using custom RegistrySharedObject and HostContainerSelector - Full Post

Hi all,

Scott tried to provide a fix for this bug. He commented on the bug that the fix should be available at


but it is not. I need version 4.2.200.qualifier of bundle org.eclipse.ecf.provider.remoteservice.

Maybe someone familiar with the ecf build could point me to the download location for this fix!?

I'm a little in a hurry for that fix, since i will be away for two weeks starting at tomorrow ;)

Thanks. Bye Peter

Am 17.03.2017 um 08:20 schrieb Peter Hermsdorf:

Hi Scott,

created https://bugs.eclipse.org/bugs/show_bug.cgi?id=513791.

I'm wondering that no one else stumbled upon this issue ....

Thanks for your help!

Bye Peter

Am 17.03.2017 um 05:14 schrieb Scott Lewis:
Hi Peter,

This seems to be a bug.   Please open a bug report with the info you've already provided, and I will fix in time for our next minor release (3.13.6).   I've already pretty much written the fix, so this can all go quick.



On 3/16/2017 4:40 AM, Peter Hermsdorf wrote:

I now removed our custom code - I don't think it has something to do with this issue.

current state:

on every client connect one ore more (depending on the number of published services) "ID targetContainerID" Objects are added to a List in the Map localRegistryUnregistrationTargets in the RegistrySharedObject on the Server Side.

the call stack is:

Daemon Thread [org.eclipse.ecf.remoteservice.IRemoteServiceContainerAdapter:run] (Suspended (breakpoint at line 1633 in RegistrySharedObject))   
    owns: RemoteServiceRegistryImpl  (id=252)   
    RegistrySharedObject.addTargetForUnregister(RemoteServiceRegistrationImpl, ID) line: 1633   
    RegistrySharedObject.sendAddRegistrations(ID, Integer, RemoteServiceRegistrationImpl[]) line: 1253   
    RegistrySharedObject.sendRegistryUpdate(ID, Integer) line: 775   
    RegistrySharedObject.handleRegistryUpdateRequest(ID, Integer) line: 1128   
    GeneratedMethodAccessor35.invoke(Object, Object[]) line: not available   
    DelegatingMethodAccessorImpl.invoke(Object, Object[]) line: 43   
    Method.invoke(Object, Object...) line: 498   
    SharedObjectMsg.doInvoke(Object) line: 323   
    SharedObjectMsg.invoke(Object) line: 290   
    RegistrySharedObject.handleSharedObjectMsg(SharedObjectMsg) line: 1843   
    RegistrySharedObject(BaseSharedObject).handleSharedObjectMsg(ID, SharedObjectMsg) line: 467   
    RegistrySharedObject(BaseSharedObject).handleSharedObjectMsgEvent(ISharedObjectMessageEvent) line: 457   
    SharedObjectMsgEventProcessor.processSharedObjectMsgEvent(ISharedObjectMessageEvent) line: 41   
    SharedObjectMsgEventProcessor.processEvent(Event) line: 36   
    RegistrySharedObject(BaseSharedObject).fireEventProcessors(Event) line: 192   
    RegistrySharedObject(BaseSharedObject).handleEvent(Event) line: 126   
    SOWrapper.svc(Event) line: 182   
    SOWrapper$2.run() line: 138   
    Thread.run() line: 745   

I don't see any code that removed entries from the List in Map when the Client disconnects. So the List (and with it the Map localRegistryUnregistrationTargets) grows until there is no memory left.

Any help on this topic would greatly be appreciated.

thx Peter

Am 15.03.2017 um 15:52 schrieb Peter Hermsdorf:
Hi all,

sorry for double posting - hotkeys are evil sometimes...

for custom error handling on server side (please see previous discussion on this list as of 27.06.2014 with the subject "Exception Handling of Remote Calls") we use a derived HostContainerSelector and a derived RegistrySharedObject. See code below.

The problem seems to be that we create a instance of RegistrySharedObject (P4RegistrySharedObject()) wich gets added to org.eclipse.ecf.provider.remoteservice.generic.RegistrySharedObject.localRegistryUnregistrationTargets but "no one" calls org.eclipse.ecf.provider.remoteservice.generic.RegistrySharedObject.sendUnregister(RemoteServiceRegistrationImpl) when the remote client disconnects.
So the Map grows over time and finally the server run's into an OutOfMemory Exception after some time.

Any suggestions where the Problem is or what is missing from the implementation?

thanks in advance!

Best Regards, Peter

BTW: we are running ECF 3.13.1.v20160405-1820

extended HostContainerSelector:

    protected IRemoteServiceContainer createRSContainer(
            @SuppressWarnings("rawtypes") final ServiceReference serviceReference, final Map<String, Object> properties,
            final ContainerTypeDescription containerTypeDescription) throws SelectContainerException {
        final ISharedObjectContainer container = (ISharedObjectContainer) createContainer(serviceReference, properties,
        final ID adapterID = IDFactory.getDefault().createStringID(IRemoteServiceContainerAdapter.class.getName());
        final ISharedObjectManager manager = container.getSharedObjectManager();
        ISharedObject so = manager.getSharedObject(adapterID);
        if (so == null) {
            so = new P4RegistrySharedObject();
            try {
                manager.addSharedObject(adapterID, so, null);
            } catch (final SharedObjectAddException e) {
                throw new P4ServerException(e);
        return new RemoteServiceContainer(container, (IRemoteServiceContainerAdapter) so);

P4RegistrySharedObject extends RegistrySharedObject and just implements a custom getSerializableException method - since the original one is private, it's a copy of the  executeRequest method.

ecf-dev mailing list
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
ecf-dev mailing list
To change your delivery options, retrieve your password, or unsubscribe from this list, visit

ecf-dev mailing list
To change your delivery options, retrieve your password, or unsubscribe from this list, visit