Job placement

Nomad scheduler defaults, node pools, preemption, and memory oversubscription are configured and managed by platform operators — see the Administration Guide: Scheduling administration.

Workload authors can influence placement by using job-level configuration that the platform team exposes for application teams.

Node pool selection

Jobs are able to opt in to a node pool by specifying the node_pool parameter. Node pool configurations override the default scheduler configuration. For example, if setting the default cluster configuration to binpack and there is a node pool with spread set, then any workloads placed on the node pool use spread.

Task resource sizing

Tasks must specify memory_max to take advantage of memory oversubscription.

job "example-job" {
  group "example" {
    task "server" {
      resources {
        cpu        = 100
        memory     = 256
        memory_max = 768
      }
    }
  }
}

To avoid degrading the cluster experience, coordinate with platform operators on resource reservations, memory limits, and resource quotas before deploying production workloads.