Commit Digest - Issue 1
4 June 2009 by Arnaud Contes
Welcome to this first commit digest post.
I thought it would be interesting for users to know what is currently going on in the ProActive Parallel Suite©. I intend to have a monthly or bi-monthly post rate depending on the number of changes that land in the repository.
For this first post, I am going to review the changes since the latest release we made last month (Programming 4.1.0 & Scheduling 1.0).
Programming:
- The implicit getstubonthis feature has been merged into the trunk. By default, an active object has to use PAActiveObject.getStubObThis() when it wants to send a reference on itself. The obvious limitation is when activating an object whose source code cannot be modified, it is not possible to replace the code that return the reference on the object (return this) by the reference on the active object.
| PROACTIVE-680 | PBS support broken due to bad escaping |
| PROACTIVE-679 | It could be usefull to have a new Property for the JMX port if used with standard tools instead of with proActive connector. |
| PROACTIVE-675 | ProActiveComponentTypeImpl.getFcInterfaceType(String) does not support collection interface name |
| PROACTIVE-673 | GCMD Descriptors: user.dir != user.home |
| PROACTIVE-672 | Router crahes after restart |
| PROACTIVE-669 | check that the provided content class of a component implements the functional server interfaces |
| PROACTIVE-665 | s/IntegerWrapper/IntWrapper/ in Programming documentation |
| PROACTIVE-663 | PAFutures.waitForAll methods require List<?> argument, should require Collection<?> instead |
| PROACTIVE-660 | wrong licence in src/Tests/functionalTests/component/collectiveitf/gathercast_remote/TestRemoteGathercast.java |
| PROACTIVE-657 | branch SHORT_ffonteno_axis2_exportation throws an exception when running under java 5 while the exception is not thrown with java 6 |
| PROACTIVE-656 | PAMR: Router.stop() does not stop the router |
| PROACTIVE-655 | PAMR: Message corruption |
| PROACTIVE-654 | GPL version is wrong in the documentation |
| PROACTIVE-644 | Remove PALauncher |
| PROACTIVE-643 | CircularArrayList.add(int,Collection) always adds to end of list |
| PROACTIVE-526 | Fill in the “1.1.5. Programming With Components: GCM” section of the ProActive manual |
| PROACTIVE-265 | synchronized all class generation methods |
Scheduling:
- Add JMX features for Scheduler and RM (1.1)
- Add visual console for controllers (1.1)
- Add visual console in Scheduler RCP (1.1)
- Improve user management (add/remove/update/authentication) by adding user temporary session (1.1)
- Fix bug in Exception thrown by javaTask (1.1 – 1.0.1)
- Improve RM node deployment (was too slow) (1.1)
- RM is client aware as it can release nodes taken by a dead client (1.1)
- Add transaction for multiple atomic changes in database (1.1)
| SCHEDULING-321 | Error at JMX level when killing a job |
| SCHEDULING-308 | In console interface, the print method arguments must not be formatted string |
| SCHEDULING-306 | attempted duplicate class definition for name in TaskClassLoader |
| SCHEDULING-304 | Node owner check (when releasing node) seems to be broken with PAMR |
| SCHEDULING-303 | attempted duplicate class definition in …scheduler.util.classloading.TaskClassLoader |
| SCHEDULING-302 | ConcurrentModificationException |
| SCHEDULING-299 | It would be great to have an historic of commands in the command line clients adminRM and adminScheduler |
| SCHEDULING-298 | Sometimes a ClassCastException occur in the Resource Manager |
| SCHEDULING-297 | SchedulerFrontend should use session to remove user that crash or disconnect without using proper disc method |
| SCHEDULING-294 | Add system independent newline characters in AdminConsole of RM |
| SCHEDULING-291 | Exceptions thrown by java tasks can be wrongly caught by non functional code. |
| SCHEDULING-290 | Resource Manager seems to be set with an incorrect address when using PAMR |
| SCHEDULING-287 | Improve sorting order for duration column in taskView (Scheduler RCP) |
| SCHEDULING-286 | Scheduler frontend should attribute session to user in order to avoid growing size of user identification list |
| SCHEDULING-283 | DataBaseManager exception occured and sended to the on client side when listening to task updated event |
| SCHEDULING-281 | Classpath under windows misses script engines |
| SCHEDULING-278 | Hide/show icons feature in a compact view of RM RPC |
| SCHEDULING-277 | NullPointerException in admin console of rm |
| SCHEDULING-276 | org.hibernate.exception.LockAcquisitionExceptionIN from logger SCHED_HIB.JDBCExceptionReporter.logExceptions in SchedulerDev.log |
| SCHEDULING-275 | Useless println in Scheduler RCP |
| SCHEDULING-274 | *Executable should not be Serializable anymore… |
| SCHEDULING-272 | SCHEDULING-270 |
| SCHEDULING-271 | SCHEDULING-270 |
| SCHEDULING-270 | Resorce manager optimizations for big number of nodes |
| SCHEDULING-268 | NullPointerException at org.ow2.proactive.scheduler.core.SchedulerFrontend.clearListene |
| SCHEDULING-266 | GCMD/Static empty node sources redeploy all nodes when some nodes are added. |
| SCHEDULING-255 | Listen log several times on the same running tasks -> lost of log information |
| SCHEDULING-253 | Test jobRecover fails. |
| SCHEDULING-226 | RM does not support client crash if client handle some nodes |
The Lab — On-going developments:
- The Distributed Service Id branch is scheduled to be merged soon. This feature allows to track a service across several active objects allowing, for instance, the detection of reentranct calls. Once this feature is merged, the component module is going to be rewritten to use this implementation instead of their previous component-only implementation.
- The debugger hotplug has also been greatly improved allowing the eclipse debugger to automatically discover and attach to remote ProActive Runtimes. We will discuss more on this topic on a future post.
- Data handling will be dramatically improved in ProActive thanks to Dataspaces: this new feature will provide a uniform Virtual File System view for every active object of an application, regardless the actual location and underneath protocol. Moreover, it will allow defining explicitly in the GCM-A file the input and output spaces for a given application.
No Comments »
No comments yet.
RSS feed for comments on this post. TrackBack URL







