Spectrum LSF
Sort by:
Trending
Status:
Not under consideration
Showing 385
1
VOTE
Threshold for over allocation in License Scheduler
Based on the license availability and the demand from certain project(s), LSF License scheduler can allocate more license to a project, then its define share of licenses. If limits are defined, it leads to under utilization and wastage of licenses...
3
Not under consideration
3
VOTE
Preserving job array conditions in multicluster
Job array conditions are widely used on our machines to manage when jobs are dispatched. Certain workflows require specific job stepping. Job array conditions aren't present in a multicluster environment. Our workaround involves submitting a job i...
2
Not under consideration
1
VOTE
Simplify the Removal of a specific select string or rusage from a running or pending job
It would be nice if we simply had the ability to bmod away a reservation without having to reconstruct the entirety of the bsub -R "blah" syntax. I would much prefer to do something like:
bmod --remove-reservation mylicense [options] jobid ...
V...
1
Not under consideration
2
VOTE
2
VOTE
In the "bresource -gl" output, there is slot number of "GUARANTEE CONFIGURED", "GUARANTEE USED" and "TOTAL USED", but no "PEN
In the "bresource -gl" output, there is slot number of "GUARANTEE CONFIGURED", "GUARANTEE USED" and "TOTAL USED", but no "PEND" slots number. We need the number of "PEND" slots for business purpose.
0
Not under consideration
2
VOTE
change in the way graceful exits are handled/reported
When LSF sends kill signal to job, based on the time job takes to complete, it sometimes return "0" exit code and sometimes non-zero. This leads to diff message being return by LSF commands. In this RFE I would like to address ::1. What exit code ...
3
Not under consideration
2
VOTE
cleanup of Old LOAD_INDEX records from lsb.events
The investigation indicates that LOAD_INDEX record is written to lsb.events when there is change (addition/deletion) in load indices. The records are not automatically cleaned and have no impact to performance unless there are a quite a lot of the...
1
Not under consideration
2
VOTE
2
VOTE
3
VOTE
Adding forking child to sbatchd can prevent significant delays and cascading problems associated with slow cgroup writes.
Platform LSF's root sbatchd daemon on the execution host gets stuck while writing to one of the cgroup subsystem. This causes the host to show unreachable status, and no new jobs will be dispatched to this host. Jobs which were already running on ...
2
Not under consideration