I am running a scripted pipeline with parallel branches each running on a separate pools of identical systems dedicated for the branch. Each pool is identified by a unique label and the agent directive in each branch specifies the appropriate label. Each node has one executor. The default selection criteria uses a fixed preferred list of nodes that will be selected in order, which leads to an imbalance of utilization where some nodes are heavily used while others are rarely used. This leads to the disks of some nodes filling quickly, and issues with some nodes are not noticed until there are issues with the others.
What is the behavior of the LeastLoad plugin in a situation where there a multiple least loaded nodes? My read of the code leads me to guess it will randomly choose a node, but I would like to confirm that behavior.
Thanks!