Summary: Known issues in GigaSpaces 6.0.2 (build 2002). Overview | Configuration, Jini Services, and Tools | Core | .NET | C++ | JMS

Overview

This list includes known issues since GigaSpaces 6.0.1 (build 1903).

See the list of new features and fixed issues in GigaSpaces 6.0.2.

Upgrading GigaSpaces 6.0.1? See this migration document.

Configuration, Jini Services, and Tools

• GS-3335 – Defining a timeout in the space URL using the RMI protocol does not take effect.
• GS-3461 – A command not found error is thrown when running gs.sh, since the toNative() method is used in the script before it is actually defined.
  This impacts partitioned-sync2backup spaces deployed on a GSC using JDK 1.4. -Xbootclasspath is needed, which is missing in this case.
  Workaround: Move the toNative() method before line 22, where it is used.
• GS-3304 – Cluster configuration is missing a <load-bal-find-timeout>.
  Workaround: This schema allows you to use the partitioned based cluster schema to change the <fail-over-find-timeout>2000</fail-over-find-timeout> element that impacts the load-balancer finder. If you are not using failover, this schema simply "fails" to an alternate dummy group.
• GS-3279 – When working on Linux with JDK 1.6.0-rc-b104, an org.xml.sax.SAXParseException is thrown.
• GS-3041 – When the space throws an OutOfMemory exception caused by clients performing large queries on the space, or by many clients querying large objects; the client needs to keep waiting for an answer from the server.
• GS-3398 – Reggie/Mahalo start without the injection of initialLookupLocators using the -Dcom.gs.jini_lus.locators, that is passed through the setenv file. In the services.config file, the initialLookupLocators is commented out, which means that you must set this attribute if you want the -Dcom.gs.jini_lus.locators to be injected into the service, so that the Reggie/Mahalo can use it for the LookupLocator Unicast lookup.
  
  If you don't uncomment this line (in com.sun.jini.reggie and com.sun.jini.mahalo), the Reggie/Mahalo services will use the default localhost IP port 4160.

  //NOTE: in order to enable the Jini Locators unicast discovery and registration, please
  //uncomment the next line and pass -Dcom.gs.jini_lus.locators=<host:port> in the command line
  //initialLookupLocators = BootUtil.toLookupLocators("${com.gs.jini_lus.locators}");

  The GSC and GSM do set the initialLookupLocators.

• GS-3405 – The DefaultConfig_ClusteredJMS.xml and DefaultConfig_DCache.xml files are loaded when creating a ClusterPolicy, instead of loading only space schema values.
• GS-3043 – The HelloWorld example throws an ExceptionInInitializerError.
  Support case ID 3281

Core

• GS-3343 – Update operation fails in local view space – when updating an Entry that is not valid in the local view, no notification is sent to the view. As a result, the original Entry might stay in the view forever.
  Workaround: Perform take and write instead of update.
• GS-3337 – A ClosedChannelException is thrown when trying to shutdown a space.
• GS-3291 – After a network disconnection of primary machine, a split-brain scenario occurs. Currently, disconnecting a network cable and re-connecting might cause data inconsistency, and split-brain scenarios.
  Workaround: After the cluster has been split into two islands, restart all members in the inconsistent island.
• GS-3262 – After performing a restart operation on a replicated space, the space doesn't recover data.
  Workaround: Restarting the container recovers the data.
• GS-3184 – Registrations for notify templates under a local transaction are also replicated.
• GS-3172 – When performing read with SQLQUery, the Before_Read filter receives templates that are digested into sub-queries.
  Support case ID: 3354
• GS-3566 – In primary-backup configuration, under LRU, Memory Management does work on the backup space. This could result in OOM exceptions.
  Support case ID: 3464
• GS-3147 – TakeMultiple is not supported with a return-only UID in local cache.
• GS-3146 – The HibernateDataSource does not support SQLQuery when loading data.
• GS-3089 – Notification listeners are not informed when the cluster is shut down.
  Workaround: Use the LeaseRenwalManager to get these notifications.
  Support case ID: 3308
• GS-3069 – FIFO is not supported across inherited classes – Entries from different inherited classes cannot be taken, read or notified in FIFO mode.
  Support case ID: 3305

• GS-2981 – NotifyDelegatorMultiplxer causes a thread leak when using FIFO.
  Support case ID: 3261

• GS-3441 – Writing a POJO to the space returns a Lease. When converting the Lease to a LeaseProxy, it is possible to perform getEntryInfo(); (this method should be public only for Entries and not for POJOs).
• GS-3391 – VM heap-size threshold usage percentage does not work.
  Support case ID 3440
• GS-3382 – Failover takes 3 minutes when the network fails.
  Support case ID 3398

.NET

• GS-3365 – SpaceFinder.find() with url=null throws a java.lang.NullPointerException.
• GS-3228 – Using any instance of ITransactionManager with .NET is not thread-safe.
  Workaround: Create the transaction in a critical section.

C++

• GS-3364 – GS_NATIVE_LIB_PATH is not included in the gs.bat file – needed for running C++ workers.
  Workaround: Add -DGS_NATIVE_LIB_PATH=%SERVICE_GRID_LIB%\native to gs.bat in the command line.
• The C++ Linux package provided with the installation is not functional. To download the functional package, click here.

JMS

• GS-3314 – When using a cluster, spaces don't bind JMS destinations other than the default destinations (MyQueue, MyTopic).
  Workaround:

1. Extract the DefaultConfig_ClusteredJMS.xml file from JSpaces.jar\config to JS_HOME\config.
2. Edit DefaultConfig_ClusteredJMS.xml to include the required Queue:

   <queues>
         <queue-names>MyQueue,TempQueue,tba.CheckerQueue</queue-names>
   </queues>

3. Redeploy the Processing Unit.
   
   After performing the steps above, the required queue should be available in the RMI registry of all spaces in the cluster.
   Support case ID 3398