2016.09 Update 10

New Features and Enhancement

Products Internal Number Case Number Description
All 7925   Symbolic links of log files, such as server.log, slave.log are changed to relative paths.
FlowTracer 7827 21382 The VOV_STDOUT_SPEC environment variable now supports @JOBNAME@ as part of the stdout and stderr filename formats, e.g.,
The jobname portion of these filenames is limited to alphanumeric characters, the underscore, hyphen and the period. Other characters in the jobname are filtered out. For jobs without a jobname, NOJOBNAME is used. Note that the vovserver must have the VOV_STDOUT_SPEC environment variable defined as above for this to take effect.
FlowTracer 7928 21547 No longer bind to the X display by default in vovsh. To bind, pass the -d option.
LicenseMonitor 7762 21280 ftlm_batch_report can take input fields from a file % ftlm_batch_report -inputFile <INFILE> and a template <INFILE> can be found at $VOVDIR/etc/config/lm/ftlm_batch_report.tm.
LicenseMonitor 7849 21424 LA will cut back unused tokens to maximize utilization. It will also try hard not to reduce allocations below currently running for higher weightage sites, thus reducing the likelihood of pre-emption on NC.
NetworkComputer 7959 21585 The -maxload option of vtk_slave_define now accepts simple expressions relative to the capacity value, represented by the keyword ’CAPACITY’. For example, -maxload CAPACITY*1.5 would set the maxload to 1.5 times the number of slots. Supported operators are: +-/* .
NetworkComputer 7535 20833 The Tcl procedure VovGetRevokeDelay can now be added and customized by redefining it in vovresourced/config.tcl under the SWD directory to allow users to customize the revoke delay to be used in vovreconciled. This allows users to have the revoke delay from their job classes override the default value of RESD(revokeDelay). The proc definition has been added to the documentation.
NetworkComputer 7714 21125 Enabled the @USER@ recipient keyword for the LongJobs health check notification procedure.
NetworkComputer 7715 21199 A limited-release capability, nc info -legacy, is provided to generate nc info output in the older format (prior to 2016.09). The behavior of this command can be customized via a Tcl file at .../local/vncinfo.config.tcl based on environment variable, username, project name, etc. by doing ’set NCINFO(legacy) 1’ under the desired conditions. Example:
if { [vtk_logname] eq "george" } { set NCINFO(legacy) 1 }
Important: This legacy feature will not be available in releases beyond 2016.09 (and updates).
NetworkComputer 7881   Fixed the ROWCOUNT field in REST API output so that it does not report double the correct row count.
NetworkComputer 7810 21343 Stop all running jobs button changed to Stop jobs for all users, and the default set to "Cancel" on the dialog window instead of "OK" (which also changed to "Yes, Stop All").

Resolved Issues

Products Internal Number Case Number Description
All 7807   Improved signal handling in vov/vw/vrt wrapper when waiting for a job to finish to increase stability.
All 7931   Fixed conflict with the -platform and -nocommon options in the batch installer so that the option order does not matter.
FlowTracer 7912 21517 LSF jobs can now have multiple "LSFmopts:*" parameters in the resource string.
FlowTracer 7953 21590 The capsule for Questa vcom is enhanced to handle files containing both entity and package declarations.
FlowTracer 7749 21260 Documentation for PJ has been updated to describe the new method of indicating a PJ is a system job.
FlowTracer 7776 21297 When an input or output is disconnected from a running job, do not invalidate the job. If a file was declared as an output of a running job during the job, and it is disconnected, propagate INVALID to the downcone of the disconnected output.
FlowTracer 7829   Prevent md5 barriers from invalidating the downcone when the timestamp changes but the content (md5sum) is still the same.
FlowTracer 7837 21356 When the FDL commands T_FINAL or J_FINAL are called, do not disconnect dependencies other than the ones previously declared in FDL (no runtime dependencies such as instrumented tools, stderr, etc, should be detached).
LicenseAllocator 7861 21445 LA will prevent reconciliation of licenses by vovreconciled if matching has been turned off for this license resource.
LicenseAllocator 7984   Fixed the problem with targets being set to 1 even when there is no demand.
LicenseMonitor 7601 20454 Added ability to specify a replacement pattern and string for host names obtained by LM parsers. This is configured by adding the following line to the SWD/config/parser.cfg file:
where PATTERN is a case-sensitive regular expression pattern, and REPLACEMENT is the string with which to replace the pattern.
LicenseMonitor 6962 20151 Use the parser time for checkouts in the remote LM parser instead of the remote checkout time, to prevent overlapping of checkout records.
LicenseMonitor 7863 21099 vovsql_load_checkouts now is able to process checkout log with handle# > limit of integer data type.
LicenseMonitor 7775 21279 Uncheck options on batch report via Web UI to disable the plot on that option, now matching the behavior of online history report. Batch report via CLI now takes boolean (1|0) value for options that do not require values before but will backward compatible if no value provided.
LicenseMonitor 7868 21447 Now the report UI is able to display users or any other drop-down list on Windows supported versions of all browsers (Chrome/FireFox/IE) even if the list contains binary characters.
LicenseMonitor 4216 21593 Added average usage statistic to the Efficiency Statistics report table.
NetworkComputer 7906 21516 Fixed issue that caused global namespace variables in the vnc_policy.tcl file to not be seen by the Tcl interpreter without adding a global call.
NetworkComputer 7383 20576 Start time, end time, and duration values are now validated in a call to vtk_slave_reserve to prevent values of 0 from being applied to a reservation. This prevents confusing reservation property entries in the /system/slaves/reservations FairShare group, if enabled. Slave reservation expirations and cancellations are now logged in the server log, even if the reservation expires while the slave is down. Slave properties, which store reservations for persistence, are now enabled by default for NetworkComputer instances.
NetworkComputer 7487 20455 Fixed issue where bash functions were not being defined correctly in snapprop.
NetworkComputer 7739 20807 Resumer jobs for jobs preempted via preemption rules that are waiting for HW now request the same slave and HW resources as the preempted job.
NetworkComputer 7937 21560 Prevent the following environment variables from being carried over from the job submission environment to the execution environment when using environment snapshots:
By doing so, conflicts in a mixed-version environment will be prevented.
NetworkComputer 7866   Cleanup stderr files generated by the SNAPPROP job environment when an error is caught by the Tcl interpreter when fetching a system property.
NetworkComputer 7880   Fixed the format of JSON output in the REST API to include commas between list elements.
NetworkComputer 7919 21520 The following changes have been made to non-legacy output of vsy, vovwhy, nc why, wx why, vtk_explain_status and related commands: slavelists will not appear as SLAVELIST-NOT-AVAILABLE with 0 slaves when running wx. For non-wx products, empty or non-existent slavelists will now appear under "Main Reason" in the first section of the output. This is an additive change; slavelist information will still appear in the "Additional Information" section as well. The redundant line "Analyzing job 12345" has been removed from non-verbose output. This makes the "Main Reason" section easier to read.
NetworkComputer 7958 21591 A slave spawned by vovelasticd should use the trigger job’s expected duration for its max life if the duration is longer than the slave max life configuration. This will prevent a deadlock situation where a slave is available to run a job, but cannot because it will exit before the job is expected to be finished.
NetworkComputer 7973   Modified job why-waiting analysis to report on cases where the job’s expected duration exceeds a slave’s maximum lifetime.
NetworkComputer 7975 21634 Enabled pre/post commands with use of -f option to nc run.
NetworkComputer 7982 21646 Changed documentation to suggest usage of the option "-rule" rather than "-hw" for multiple constraints in a "nc hosts" command.
NetworkComputer 5725 20830 The "Seamless Transition To A Cycle-Based Scheduler" of the docs has been modified to provide added description for scheduler parameters.
NetworkComputer 7644 21092 Improved error handling in LSF emulation’s bsub
NetworkComputer 7930 21518 Do not attempt to kill a non-existent subslave (with a PID of 0) after a job fails to start due to an unknown user error.
NetworkComputer 7631 21020 If job class description has html tag <em>, the enclosed text will rendered to the intended effects as italic/emphasized.
NetworkComputer 7682 21153 Fixed so that all NC jobclasses from $VOVDIR/local/jobclass show up in the web GUI.
WorkloadAccelerator 7890 21478 Enabled initialization of jobclasses so that jobclass limit resources are defined correctly.
WorkloadAccelerator 7910 21532 Made vovelasticd sensitive to bucket FairShare groups for determining the number of job launchers to submit to the back-end queue.
WorkloadAccelerator 6281   Added "VOVELASTICD(maxQueueErrors)" setting, defaulting to 10, that will cause the vovelasticd daemon to stop attempting launcher submissions when the number of consecutive launcher submission failures to the back-end queue exceed the setting. Both a system alert, as well as a vovelasticd log file entry, will be generated during this condition. Most often, this is related to a misconfigured queue. The daemon will resume normal operation once the configuration files have changed to correct the issue.
WorkloadAccelerator 6314   Documented the requirement to include LSFqueue:<queueName> in the resource specification in order to direct jobs to a specific LSF queue when using LSF as the base scheduler.
WorkloadAccelerator 7744 20975 Enabled additional health check procedures, which can be used to monitor for errant conditions in the compute environment or the workload, such as jobs that have been queued or running for longer than desired.
WorkloadAccelerator 7772 21266 Failover slaves now use the original vovserver’s VOVDIR as opposed to their own when starting a failover server.
WorkloadAccelerator 7820 21173 Fixed case in WX where users first jobs do not run when they are submitted with "-f file".
WorkloadAccelerator 7969 21614 Slaves spawned by vovelasticd should be reserved for the trigger job’s bucket instead of its FairShare group. This will ensure that the slave will execute the trigger job first, then any subsequent jobs in the same bucket instead of allowing unrelated jobs to be executed.
WorkloadAccelerator 7976 21633 Elastic slave launchers should reflect the job placement policy of the trigger job.
WorkloadAccelerator 7989 21661 Elastic slaves should be initialized with actual RAMTOTAL instead of using the trigger job’s requested RAM.