compute_tpl_8_cpu_autoscaling.yaml 875 B

12345678910111213141516171819202122232425262728293031323334
  1. cloud_id: {{env["ANYSCALE_CLOUD_ID"]}}
  2. region: us-west-2
  3. # 1k max replicas (1000 / 8 = 125 containers needed)
  4. max_workers: 130
  5. head_node_type:
  6. name: head_node
  7. # 8 cpus, x86, 32G mem, 10Gb NIC, $0.384/hr on demand
  8. instance_type: m5.2xlarge
  9. resources:
  10. custom_resources:
  11. proxy: 1
  12. worker_node_types:
  13. - name: worker_node
  14. # 8 cpus, x86, 32G mem, 10Gb NIC, $0.384/hr on demand
  15. instance_type: m5.2xlarge
  16. # smaller min workers will make the head node cpu usage very high, and crash the head node.
  17. # issue: https://github.com/ray-project/ray/issues/18908
  18. min_workers: 5
  19. # 1k max replicas
  20. max_workers: 130
  21. use_spot: false
  22. resources:
  23. custom_resources:
  24. proxy: 1
  25. aws:
  26. TagSpecifications:
  27. - ResourceType: "instance"
  28. Tags:
  29. - Key: ttl-hours
  30. Value: '24'