Prototype of Persistent Flux MiniCluster

by vs
GNU/Linux ◆ xterm-256color ◆ bash 597 views

Instead of creating a new cluster per job, we instead create a MiniCluster of one size and then submit jobs to it via the Flux RESTFul Python API. That connection is likely the slowest bit - needing to query for info and get logs, that maybe we can improve upon.