Vnode_per_numa_node and custom resource queue_list

Users/Site Administrators
1

Hi there,

I am trying to install a OpenPBS over RockyLinux. There are two servers and a client (VMs with RL 8.6) and a MoM (physical machine with RL 8.7; I can put 8.6 but not immediatly; installation of the OS was not in my control).

I am using the pbs_cgroups hook with "vnode_per_numa_node": true and "use_hyperthreads" :false. The node as 52 physical cores and 4 GPUs and is configured with the custom resource queue_list which is rather a common way to associate nodes to queues (as the AG says):

# Create and define resource queue_list
create resource queue_list
set resource queue_list type = string_array
set resource queue_list flag = h
set queue workq default_chunk.queue_list = smalljobs
set queue bigq default_chunk.queue_list = bigjobs
[root@pbs-test03 ~]# grep bigjobs node05.qmgr
create node pbs-test05 resources_available.queue_list = bigjobs

I’d expect that the resouce would be copied in the vnodes created when I enable the hook and then restart the MoM but this is not the case, pbsnodes -av shows resources_available.queue_list = bigjobs only for the host. I have tried to recreate the node(s) a few times.

While it’s easy to fix I wonder if I am missing something. There’s a way to automatically import resources?

thank you

2

No you are not missing any configuration. Your way of configuration is correct , manualy configure the custom resources on the vnode created by cgroups.
You can use the v2 config file, V2 config questions - #4 by agurban

3

Hi Adarsh,

thank you so much for your help. I will check out V2 conf files.