News and Information

Notification on Changes to the Default Execution Queue for UGE Jobs

To all users,

Thank you for using our center's services.

In order to prevent errors due to differences between the CPU expanded instruction functions of compute node Thin and computation node Fat, as well as to ease congestion resulting from jobs requiring a lot of memory, from the following date and time, the operation of UGE will be changed so that jobs that are qsub-ed without any queue options, will not be executed by lmem.q.

06/22/2018 15:00

Jobs that are qsub-ed and awaiting execution prior to this date and time will follow the previous rules.

In future, when using computation node Fat (lmem.q), you will need to append "-l lmem" in the qsub options.

Ex. $ qsub -l s_vmem=256G,mem_req=256G -l lmem [jobscript name]

Just as before the changes, if you want to add lmem.q as a candidate to execute a job, please declare "-soft -l -lmem" as in the example below.

Ex. $ qsub -l s_vmem=100G,mem_req=100G -soft -l lmem [jobscript name]

Additionally, in order the minimize the effect of these changes, the system will at the same time be changed so that qsubs of jobs, the memory demand of which is indicated as exceeding 128GB (mem_req), will automatically have "-l lmem" appended to them if they do not already declare "-l lmem."

Summary of Changes
When using computation node Fat (lmem.q), you need to append "-l lmem" to the qsub options.
Jobs that have previously been qsub-ed and executed will continue to execute unchanged.
While there may be cases where time to execution is extended, correct appending of options should allow this to be avoided.
By declaring "-soft -l lmem" at the end of qsub options, jobs will be executed according to the same settings as before these changes.
If memory demands are in excess of 128GB, please declare the above or "-l lmem."

Please contact us if you have any points you need clarification on.

Index

Top of Page Top of Page

The University of Tokyo The Institute of Medical Science

Copyright©2005-2018 Human Genome Center