Limit one job per node


#1

Hi team,

please help, how to configure PBS Pro for allow only one job per node ?

Regards
Pradeep


#2

#3

Per node level you can

qmgr -c "s n nodeA max_running=1"

or as Adarsh suggested use resources request on job script

-l place=excl

#4

Reason for not suggesting max_running attribute is because :
18.x.x End of support The max_running vnode attribute is no longer supported


#5

A quick test from my environment

$ qsub --version
pbs_version = 18.1.0
$ pbsnodes debian
debian
     Mom = debian
     ntype = PBS
     state = free
     pcpus = 4
     jobs = 17994.debian/0
     max_running = 1
     resources_available.arch = linux
     resources_available.host = debian
     resources_available.mem = 7994652kb
     resources_available.ncpus = 4
     resources_available.vnode = debian
     resources_assigned.accelerator_memory = 0kb
     resources_assigned.hbmem = 0kb
     resources_assigned.mem = 0kb
     resources_assigned.naccelerators = 0
     resources_assigned.ncpus = 1
     resources_assigned.vmem = 0kb
     comment = abc
     resv_enable = True
     sharing = default_shared
     last_state_change_time = Fri Sep 14 18:39:49 2018
     last_used_time = Wed Sep 12 17:46:04 2018

$ qstat -ans$ qstat -ans

debian: 
                                                            Req'd  Req'd   Elap
Job ID          Username Queue    Jobname    SessID NDS TSK Memory Time  S Time
--------------- -------- -------- ---------- ------ --- --- ------ ----- - -----
17994.debian    source   workq    STDIN       16566   1   1    --    --  R 00:01
   debian/0
   Job run at Fri Sep 14 at 05:13 on (debian:ncpus=1)
17995.debian    source   workq    STDIN         --    1   1    --    --  Q   -- 
    -- 
   Not Running: Node has reached job run limit
17996.debian    source   workq    STDIN         --    1   1    --    --  Q   -- 
    -- 
   Not Running: Node has reached job run limit
17997.debian    source   workq    STDIN         --    1   1    --    --  Q   -- 
    -- 
   Not Running: Node has reached job run limit

it still works at least for 18.1.0, does it mean that it still works, but without any warrant?


#6

max_running, max_group_run and max_user_run on vnodes were deprecated in 14.X.X. Some code optimisations in the scheduler do not inter-operate correctly with them. It is mentioned in the release notes and is mentioned in the 18.x.x guides.