Google Cloud VM Reservation Not Utilized by Kubernetes Cluster Node in Same Zone

[yohann84L]

I'm encountering an issue with a VM reservation in Google Cloud not being utilized by a Kubernetes cluster node, even though they are in the same zone. Below are the details of my setup:

• Kubernetes Cluster Configuration:
  • Node Pool named "gpu" with autoscaling enabled, spanning across three zones: europe-west1-b, europe-west1-c, and europe-west1-d.
  • Each node is configured with the following specs:  n1-highmem-2 machine type, 1 x NVIDIA T4 GPU.
• Reservation Details:
  • Active reservation in europe-west1-b.
  • Reserved for n1-highmem-2 machine type with 1 x NVIDIA T4 GPU.
• The issue is that the reservation shows 0 for "Machines Utilized", despite having a node in the europe-west1-b zone which matches the reservation specs. This node was created after the reservation was active.

I have verified the following:

• The reservation and the node have matching configurations (machine type, GPU).
• The zone of the reservation (europe-west1-b) is one of the zones in the node pool.

I am unsure why the reservation is not being applied to the node in the same zone and would appreciate any guidance on what might be causing this issue or any additional checks I should perform.

Thank you for your help!

[lawrencenelson · Accepted solution]

Hi @yohann84L,

Welcome to the Google Cloud Community!

Please check if the --reservation-affinity flag of the VM is configured properly. You may follow this documentation. Also, consider whether proper taints and tolerations are configured.

Please let me know if that was helpful. Thank you! 😃