Config for node_sort_key won't work

Users/Site Administrators
1

edit sched_config as:

node_sort_key: “ncpus LOW assigned” ALL
# node_sort_key: “sort_priority HIGH” ALL

Qmgr: set server scheduling = False

killl -HUP <pbs_sched pid>

Qmgr: set server scheduling = True

but when qsub job, all job going to be run at same node, the frist node:

cat test.o*
begin at 11:30:07 CST @node1
begin at 11:30:07 CST @node1
begin at 11:30:07 CST @node1
begin at 11:30:07 CST @node1
begin at 11:30:08 CST @node1
begin at 11:30:08 CST @node1

If systemctrl restart pbs needed?

pbs_version = 19.1.3
(on a really old centos cluster)

2

Not required. Your steps are correct.
kill -HUP < pid of pbs_sched >
qmgr -c “set server scheduling = t”

Try to run jobs that run for a longer time (say 3 minutes) .

3

node1 and qq0 is config to run long time runing task only (that’s why I’m not just do a reboot), and not used by qq;

and every new jobs submit to qq just got to run on node2;
thare are 5 nodes with exactly the same hardware, and several nfs/oss servers;

only when node2 comsume all resouce, would job be run on next node and so on;

high peak of jobs on single node, just overload nfs traffic on this node;
that is annoying;

mgr:
Req’d Req’d Elap
Job ID Username Queue Jobname SessID NDS TSK Memory Time S Time

145036.mgr dummy qq0 xxxxxxxxxx 4496 1 1 – – R 461:5 node1/0
150901.mgr dummy qq0 xxxxxxxxxx 4497 1 1 – – R 461:5 node1/1
163041.mgr dummy qq0 xxxxxxxxxx 4498 1 1 – – R 461:5 node1/2
193511.mgr dummy qq xxxxxxxx 24761 1 1 – – R 78:59 node2/1
197818.mgr dummy qq0 xxx 11130 1 1 – – R 05:19 node1/3
197819.mgr dummy qq xxxxxxxxxx 40684 1 1 – – R 05:19 node2/0
197820.mgr dummy qq xxxxxxxxxx 40718 1 1 – – R 05:19 node2/2
…

All other config should be left default;
I would try a systemctrl restart later, to check if that works;

thank for you help anyway !

4

Yes it worked.

I check my config file, and find out that there is a misplaced Non-ascii character;
It turn out to be the configfile is broken.

1 Like