=============
OAR CHANGELOG
=============

version 1.6.2:
==============
    - debug reservation. Forbid reservation to finish after the initial stop time.
    - add tag in oar.conf "DEAD_SWITCH_TIME" that specify the time after which Absent and Suspected nodes will become Dead

version 1.6.1:
==============
    - bug fix about sudo : since 1.6.8p12 sudo does not propagate PERL5LIB variable
    - initialise the "ganttJobsPrediction" table with a right reference date (1970-01-01 00:00:01)
    - oarsub has the "-k, --checkpoint" option. It specifies the number of seconds before job walltime to send a SIGUSR2 on it.
    - You can see the list of events for jobs with -f option in oarstat
    - oardel can now send SIGUSR2 to the oarexec of a job (-c option)
    - oardel can now delete several jobs
    - Add a signal handler in oarexec and propagate this signal to the user process
    - Support '#OAR -p ...' options in user script
    - Add in oar.conf the possibility to specify a different timeout for prologue and epilogue (PROLOGUE_EPILOGUE_TIMEOUT)
    - when a connection to the DB fails, OAR retries 5 times
    - handle the EXTERMINATE_JOB event and suspect all nodes if possible
    - add job state log table
    - add node properties log table
    - add possibility to use sentinelle.rb script in the ping_checker module
    - add -s option to oarnodes command --> show only state of nodes
    - add -l option to oarnodes command --> show only the node list
    - root can now delete any jobs in the same way the oar user could
    - change a few oardel messages
    - limit the commands (with arguments) specified by users::

        regular expression : [\w\s\/\.\-]*

    - change oaremovenode command to oarremovenode
    - enhance job error management
    - enhance suspicious nodes detection
    - fix bugs about accounting

version 1.6:
============
    - debug reservation jobs (if the job has not the right number of requested nodes then it will wait for a delay. Once expired, the job will be launched on the nodes available).
    - add a cache for visualization of the gantt scheme (add two gantt table for the visu)
    - add an event log mechanism. It permits to know all decisions and events occuring in OAR with regards to jobs.
    - detection of errors (they can be traced via event_log table):
        * job working directory does not exist on the node
        * output files cannot be created
    - add deploy scheduler awareness (schedule on non deploy nodes firstly)
    - possibility to change property value via oarnodesetting command (-p option)
    - add the command "oaremovenode" that allows to delete a node of the database definitely
    - bug fix : now oarsub can use the user's script even if the oar user cannot
    - now, you can use special value "all" in oarsub ("nodes" resource). It gives all free nodes corresponding to the weight and properties specified.
    - debug Gantt visualization
    - add the possibility to test nodes via nmap
    - add accounting features (accounting table)
    - change nodeState_log table scheme to increase rapidity

version 1.5:
============
    - internal development version

version 1.4:
============
    - add "DEPLOY_HOSTNAME" tag in oar.conf
      This variable specifies the node where OAR will be connected to submit a job in the deploy queue.
    - "besteffort" jobs are now killed by the oar_meta_sched. This feature simplifies the writing of new schedulers as they do nothave to care about "besteffort" jobs.
    - bug fix about job startTime and stopTime (no more overlapping)

version 1.3:
============
    - Merging of the desktop computing branch.
    - Add some changes for the kaDeploy mode (for a kaDeploy reservation, you are logged on the server instead of the first node).
    - Add a new module "finaud" : check every 5 minutes if nodes in state "Suspected" are still suspected and if the ones in state"Alive" are still alive.

version 1.2:
============
    - change ping checker default strategy - do 10 pings in root.

version 1.1.2:
==============
    - bug fix on stopTime affectation in bipbip

version 1.1.1:
==============
    - bug fix on oarexec waitpid

version 1.1.0:
==============
    - bug fix (MySQL database deconnection)
    - oarexec can be detached from the server. So OAR is now more scalable and the server can be rebooted without risking losing the running jobs. To enable/disable this feature, look for the DETACH_JOB tag in the oar.conf file.

version 1.0.0:
==============
    - 0.9.7 version released as the stable version
    - Fix CHANGELOG :-)

version 0.9.7:
==============
    - Complete change of installation procedure (Makefile)
    - Debian packages are now available
    - Examples of prologue/epilogue that can clean SMP better are provided
    - Use of a sudo wrapper instead of setuid bit on perl scripts (now users must be in oar group)

version 0.9.6:
==============
    - Socket connection between server and node is removed (this socket permitted to send kill notification to oarexec. Now you can kill oarexec directly)
    - User commands are launched by means of a wrapper. It executes sudo command and set variables for the perl library and OAR installation paths. You must configure sudo for oar group (there is no longer a suid bit in the new installation).

version 0.9.5:
==============
    - Add reservation feature
    - Bug fix
    - Scheduling policy templating has been re-enabled for each of the queues.

version 0.9.4:
==============
    - Gantt scheduling notion in scheduler architecture

version 0.9.3:
==============
    - bug fix
    - change kill method of Leon

.. $Id: CHANGELOG,v 1.41 2005/10/03 14:36:36 capitn Exp $
