Moab 6.1.5
 
- Fix for configure script incorrectly setting MAX_ATTRS=$MAX_GRES
- Fix for corrupted job completion record in checkpoint file causing 
moab core dump
- Fix for preemptors running outside of their reservation
- Fix for preemption on inaccessible resources
- Fix for msub -v to handle variable names with commas
- Fix to display correct GRes usage on job startup.
- Fix so that showq does not display completed jobs in the idle queue
- Fix error message for script exceeding buffer size
- Fix to display correct PE for Cray with OpenMP
- Fix for display of limits in mdiag L -v
- Fix to remove unnecessary midag -C alert messages
- Fix some incorrect error messages displayed by checkjob in the grid
- Fix to control class hostlist in a cray environment.
- Fix cray script (job.query.xt4.pl) to support multi-req jobs
- Fix for jobs going deferred after a moab restart
- Fix for reported error "cannot submit job - maximum job count reached" 
when JOBCPURGETIME set to 0
- Fix for various segfaults
- Fix for multi-req cray jobs to get correct nodecount set.
- Fix for nodeset spanning

- Enable additional checkjob messages to report nodeset failures due to 
inadequate resources
- Enable a node allocation policy for reservations RSVNODEALLOCATIONPOLICY
- Enable increased size for severl Moab internal data buffers
- Enhancement to job.start.xt4.pl script
- Enhancement for NodeAllocationPolicy Contiguous in general and for Crays.
- Set CHECKSUSPENDEDJOBPRIORITY to TRUE by default
- Enable links on client commands
- Enable wrapping -l hostlist line in job.query.xt4.pl script for Cray 
submissions.
- Enable configuration of a default wclimit for users USERCFG[DEFAULT] 
DEFAULT.WCLIMIT=xxx


Moab 6.1.4

- Added "description" and "variables" attributes to cloud environment 
pending action xml
- Added additional nodeset information to checkjob -v -v output
- Fix a problem with interactive jobs in a grid using the global gres
- Fix a problem with job wallclocktime when the wallclocktime is changed 
(via mjobctl) while the job is idle
- Fix a problem where moab was not recognizing when a job dependency was 
removed in Torque
- Fix a problem where msub did not properly handle job scripts when 
submitting interactively
- Make the default for the GUARANTEEDPREEMPTION configuration option to 
be TRUE
- Enable mdiag -j ALL option displays completed jobs and system jobs
- Enable modification of the VM state with mvmctl -m
- Enhanced support.diag.pl to package "#INCLUDE" files, create tarball 
in /tmp directory, and include checknode output
- Replaced the --with-gold configure option with --with-am to support 
Moab Accounting Manager 


Moab 6.1.3

- Added new parameter 'HYPERVISOR' to vmstorage job triggers to support 
block storage requirements.
- Added 'sched' attribute to pending action XML.
- Enable automatically retrying (3 times per iteration) to connect to DB
- Enable "mrmctl -R PAR:<name>" to reload per partition config info
- Enable storing variables in description
- Enable Moab to use email address specified with qsub -M
- Enable capability to set file permissions for log files
- Enable per-job JobRejectPolicy
- Enable force cancel on generic,vmstorage,vmdestory,migration jobs.
- Enable multiple classes in the class: expression for reservation host 
list.
- Enable mnodectl to erase variables.
- Hypervisors with 0 load are not considered as migration targets
- Improved gpu integration with Torque. Supports the new 'qsub -l 
ncpus=x:gpus=y' syntax.
- Added "features" to MVMChooseVLANFeature
- cray size 0 jobs supported with gres
- Make GRESTOJOBATTRMAP case insensitive
- Sort active queue in showq by remaining time
- Fix for negative AProcs in mdiag -n
- Fix for reporting completed job walltimes
- Fix for looking up moab.dat
- Fix for preemptors preempting resources on inaccessible resources
- Fix for nodesets within a grid.
- Fix for showq for HIDEBLOCKED jobs
- Fix column alignment in showq -v outputs with large usernames.


Moab 6.1.2 Changes

- Enable RMCFG[] FLAGS=FlushResourcesWhenDown for license managers
- Enable job modify to force a checkpoint regardless of LimitedJobCP flag.
- Changed NodeSetPriorityType default from MINLOSS to FIRSTFIT
- Fix for submitting jobs with "msub -l ncpus="
- Fix for better walltime synchronization between Moab and Torque
- Fix for showing completion code in "showq -c"
- Fixes for scheduling GPU jobs
- Fixes for NUMA scheduling


Moab 6.1.1 Changes

- Enable SCHEDCFG[] FLAGS=AlwaysUpdateCache (MOAB-2866)
- Enable sharemem support for multiple nodesets.
- Enable out-of-band VM modification
- Fix for rejecting invalid syntax in "mschedctl -m"
- Fix for CLASSCFG[] HOSTLIST
- Fix for pruning old job checkpoint entries
- Fix for GPU charge rates with GOLD
- Fixes for VM migrations