Spectrum LSF

Showing 488

Multi-cluster pre-exec failure should kill job from source cluster

These documents make it look like this issue is addressed, but they are not functioning as they should. https://www.ibm.com/docs/en/spectrum-lsf/10.1.0?topic=lsbparams-remote-max-preexec-retry https://www.ibm.com/docs/en/spectrum-lsf/10.1.0?topic=...

about 2 years ago in Spectrum LSF / Scheduling 0 Delivered

4 VOTE

Prevent LSF from dispatching jobs to problematic hosts consistently

Some jobs are dispatched to a host and failed because the communication between MBATCHD and SBATCHD is failed. The host keeps in OK status and the jobs keep dispatching to the host, which makes the job can not run for a long time. The communicatio...

about 9 years ago in Spectrum LSF 2 Delivered

3 VOTE

if there is not job tmp directory, RES will re-create it

To have better control over RES created files like .1513002449.181663317.acct , which are responsible for proper accounting of HPC jobs. Currently these intermediate files are created under directories defined through LSB_JOB_TMPDIR or LSF_TMP_DIR...

almost 7 years ago in Spectrum LSF 0 Delivered

2 VOTE

Prevention of Spectrum LSF over-preemption

The current LSF preemption design does not prevent over-preemption (i.e. preempt more jobs than needed). When the preemption of a large low job takes some time like several minutes, the following incoming high priority jobs could preempt other run...

over 4 years ago in Spectrum LSF 1 Delivered

2 VOTE

Use coherent units and unit abbreviations for memory sizes in `bjobs -o` output

bjobs -o "max_mem memlimit" creates a hard-to-parse-and-process output because it a) doesn't respect LSF_UNIT_FOR_LIMITS for the numbers b) uses different units for the numbers c) uses different abbreviations for the same unit ("Gbytes" vs. "G") T...

over 4 years ago in Spectrum LSF 1 Delivered

4 VOTE

many jobs pend due to "Just started a job recently".

1. the candidate hosts are sorted by CPU load; 2. do not dispatch jobs in a "pack" fashion, instead, "balance" fashion.

over 9 years ago in Spectrum LSF 1 Delivered

4 VOTE

Support resizable jobs for compound resource requests (RES_REQ) using blaunch (parallel jobs)

A job that has compound resources cannot be resized - why?=============================================== We use blaunch to launch jobs with the following resource requirements. Begin ApplicationNAME = core8DESCRIPTION = applications that need 1 H...

over 9 years ago in Spectrum LSF 3 Delivered

2 VOTE

Take job size into account in historical run_time

LSF's fairshare includes a RUN_TIME parameter that can track the total time accumulated by a user / group's completed jobs. By weighting this parameter, a user's historic jobs can be accounted for in determining their job priority. This factor doe...

almost 5 years ago in Spectrum LSF 2 Delivered

3 VOTE

Jobs that get into PSUSP state because they reached MAX_JOB_REQUEUE cannot be killed by an appropriate JOB_CONTROL mechanism

Answer from PMR: For the PSUSP jobs JOB_CONTROL can not work on PSUSP job. We have checked the source code in these days and found that 1. LSF doesn't set any environment variable when a job reaches its requeue limit and be put into PSUSP status 2...

about 7 years ago in Spectrum LSF 1 Delivered

4 VOTE

Remove string length restrictions on resource requirement string to bsub

Our job submissions use quite complicated compound resource requirements. As our clusters increase in complexity we find our resource requirement strings used to bsub are becoming too long for the LSF limitation of 511 characters.

over 9 years ago in Spectrum LSF 2 Delivered

Please enter your email address

FILTER BY CATEGORY