How to run long time jobs (process) at RARFAXP

last updated on 18 June 2003 by T.Ichihara

Batch Queueing System (LSF) has been installed on RARFAXP, RARFAX1, and RARFAX2 unix servers. Please use LSF Batch Queueing System to execute jobs(processes) which exhaust CPU time more than 1 hour. Do not run more than two jobs at RARFAXP. Otherwise the jobs (processes) may be canceled.

[Example 1]

Assuming a script for a job as "run.sim", as follows.

(file run.sim)

#!/bin/csh
/rarf/u/myname/super/sim.bin
exit

To submit this job, type as follows.

bsub -q axp_long -o sim.o.%J -e sim.e.%J run.sim

Then your jobs will be submitted and result will be written in a file sim.0.xxxxxx and sim.e.xxxxxx (xxxxxx is job number)

[Example 2]

If you want to execute your jobs on RARFAXP (not on RARFAX1 or RARFAX2), then submit the jobs as follows.

bsub -q axp_long -m rarfaxp -o sim.o.%J -e sim.e.%J run.sim

There are three queues, axp_short, axp_long and axp_bg. Maximum CPU time is limited to 2 hours, 24 hours, 7-days, respectively.

Related LSF command (see help by "man command-name")
lsid display the LSF name and current LSF master node
lsload display load average of LSF job node
bsub submit a job to LSF batch queueing system
bkill kill a LSF job
bswitch change queue of a job before execution
bjobs display the executing jobs
bjobs display the executing jobs
bqueues display the queue states
xlsf start the LFS GUI interface
lsrun execute command on free-cpu node
lstch tcsh for LSF
bhosts display host status

RARFAXP UNIX Cluster LSF queue setting

Queues property

queue-name
Max CPU time
nice-value
hosts
Total Jobs
axp_short
2 hours
1
rarfaxp -ax1 -ax2
5
axp_long
24 hours
10
rarfaxp -ax1 -ax2
4
axp_bg
7 days
19
rarfax1 -ax2
2

Max jobs for each jobs

rarfaxp:0 rarfax1:3 rarfax2:3

priority of the queue axp_short = axp_long > axp_bg

Max executing jobs per person : 2


last-updated on 18 June 2003