Large numbers of nodes and qmgr set vs create node

in our current environment nodes only exist when a job is requesting them, and go away when the job is done. So an actual qmgr create node command is run when it starts up and a qmgr delete node happens when the job is gone.
The create node call is (in computer terms) slow, a bottle neck level slow.
(in human terms its ~0.4 seconds)
the alternative is to pre create 250K+ nodes once when the facility is created and then call a set node to update info on the node instead of a create node command when a job is submitted.
so my question here is in 2 parts,

does having 250K nodes that are mostly down/offline slow down the PBS server or scheduler to any noticeable extent?

does anyone suspect that a qmgr set node command will happen significantly faster than a create node command?



No, there would not be any permformance issue with server or scheduler. But, if you have any use case request to share it .

You copy the config_v2 file to $PBS_HOME/mom_priv/config.d/ folder before pbs_mom is started up on the compute nodes. Hence, you can avoid qmgr set node command all togheter.

the config.d directory concept looks promising. I hadnt ever seen it in the docs before.
but going the preallocate generic nodes and updating when the mom starts should be
way faster.


Thank you @steveheistand
Please check this section 3.4.3 Version 2 Vnode Configuration Files
in this document: