Jump to content
  • Announcements

    • admin

      PBS Forum Has Closed   06/12/17

      The PBS Works Support Forum is no longer active.  For PBS community-oriented questions and support, please join the discussion at http://community.pbspro.org.  Any new security advisories related to commercially-licensed products will be posted in the PBS User Area (https://secure.altair.com/UserArea/). 

How to prevent from arranging particular nodes work together

Recommended Posts



There are 22 nodes,  the high delay issue only happened when job  crossed C19 & C21,  but there was no high delay issue when job crossed C19 with other nodes except C21,   this was same to C21.   In other words,  either C19 or C21 can work with other nodes perfectly,   just only have high dealy issue between C19 & C21 each other. 


So I just wondering if it is possible to prevent from arranging C19 and C21 together but allow to arrange C19 or C21 with other nodes. So it still can have full working nodes instead making C19 & C21 offline .


Thanks for your time



Share this post

Link to post
Share on other sites

It is interesting that there is a high-delay between your nodes - especially if they are all the same. Sounds like a network switch/setting that could be optimized. IMHO, I would make sure you understand why there is a delay between the nodes before trying to implement a scheduling policy to workaround the issue. Okay, that said.. Have you considered using placement sets (PBS Professional AG v12.2, Section 4.8.32 Placement Sets)?


You could create two placement sets where C19 and C21 can only exist in one of the groups. 


group    type=string_array  flag=n


qterm -t quick ; /etc/init.d/pb start


qmgr -c "set server node_group_enable = true"

qmgr -c "set server node_group_key = group"


for nodes C00-C18, C20

qmgr -c "set node <nodename> resources_available.group = 'blue, green'"


for node C19

qmgr -c "set node C19 resources_available.group = blue"


for node C21

qmgr -c "set node C21 resource_available.group=green"


As you will find in the AG, the user will not need to request the group resource at submission. The node_group_key will be information the scheduler will use when placing the job. 


If you want to let the users request a specific group, the AG will discuss that setup, too. 

Share this post

Link to post
Share on other sites



Thanks for the quick response.



It's a wiered hardware issue,  made a replacement,  the problem was still not fixed.   The workaound is just not making C19 and C21 work together.   Your instruction are clear and good, it looked like address my problem.


Appreciate your time





Share this post

Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in

Sign In Now