all: children: cluster: children: manager: hosts: compute000: workers: children: compute: hosts: compute003: gpus: hosts: compute002: compute004: #compute005: vars: single_node: false manager_ip: 10.0.0.100