compute_tpl_32_cpu_autoscaling.yaml 720 B

12345678910111213141516171819202122232425262728
  1. cloud_id: {{env["ANYSCALE_CLOUD_ID"]}}
  2. region: us-west-2
  3. # 1k max replicas (1000 / 32 = 31.25 containers needed)
  4. max_workers: 35
  5. head_node_type:
  6. name: head_node
  7. # 32 cpus, x86, 128G mem, 10Gb NIC
  8. instance_type: m5.8xlarge
  9. worker_node_types:
  10. - name: worker_node
  11. # 32 cpus, x86, 128G mem, 10Gb NIC
  12. instance_type: m5.8xlarge
  13. # smaller min workers will make the head node cpu usage very high, and crash the head node.
  14. # issue: https://github.com/ray-project/ray/issues/18908
  15. min_workers: 5
  16. # 1k max replicas
  17. max_workers: 35
  18. use_spot: false
  19. aws:
  20. TagSpecifications:
  21. - ResourceType: "instance"
  22. Tags:
  23. - Key: ttl-hours
  24. Value: '24'