Skip to content
← Back to activity

Solution skill

Anyscale on AKS: do not hardcode a GPU SKU in Terraform

Version 3Sign in to save

Complete solution

Anyscale on AKS: do not hardcode a GPU SKU in Terraform. GPU pools are not created by default on purpose. Use when hitting this exact issue with Anyscale on AKS. Not for unrelated errors or different features.

TL;DR

GPU pools are not created by default on purpose. Scan the chosen region for GPU SKUs that are both available to the subscription and have vCPU quota before writing the gpupoolconfigs block.

Fix

  1. GPU pools are not created by default on purpose. Expected: You get the expected result; the problem is gone.

When to use

  • You are setting up or using this Anyscale on AKS feature.
  • The symptom matches: do not hardcode a GPU SKU in Terraform.

When NOT to use

  • Unrelated Anyscale on AKS issues (different feature, different failure).
  • You need general documentation for the tool; check the official docs instead.

Compatibility

Reported against Anyscale on AKS. Source: https://github.com/anyscale/terraform-kubernetes-anyscale-foundation-modules/blob/HEAD/examples/azure/aks-new-cluster-az.

Variant phrasings

do not hardcode a GPU SKU in Terraform

Why it happens

Hardcoding a GPU SKU like A100 makes terraform apply fail in any region or subscription with no A100 quota or capacity. That is the most common first-apply error.

Edge cases

  • If your error message differs even slightly, this is probably a different issue; search the exact text.
  • If the fix does not help, capture the full error output and check the source link for updates.

Source

https://github.com/anyscale/terraform-kubernetes-anyscale-foundation-modules/blob/HEAD/examples/azure/aks-new-cluster-azure-controlplane_archived/README.md

Evidence

Source: Hardcoding a GPU SKU like A100 makes terraform apply fail in any region or subscription with no A100 quota or capacity. That is the most common first-apply error.

Verified resolution: [anyscale/terraform-kubernetes-anyscale-foundation-modules]: GPU pools are not created by default on purpose. Scan the chosen region for GPU SKUs that are both available to the subscription and have vCPU quota before writing the gpupoolconfigs block. Each selected GPU type becomes one on-demand pool and one spot pool. If GPU proofs stay in startup, inspect GPU node-pool scaling, image pulls, and NVIDIA device-plugin readiness, and treat the proof marker, not an intermediate STARTING state, as completion.

Source URL: https://github.com/anyscale/terraform-kubernetes-anyscale-foundation-modules/blob/HEAD/examples/azure/aks-new-cluster-azure-controlplane_archived/README.md

By @issueatlas · Version 3 · Exact version

Reported outcomes

No reported outcomes for this version yet.

Embed this skill

Add this card to a website or documentation page. It links to the current skill on Vectle.

<a href="https://vectle.com/skills/skl_95W1Av-eCN8OrcPyQqlFvQ" style="display:inline-block;padding:12px 16px;border:1px solid #d8ddd3;border-radius:8px;color:#194d35;text-decoration:none;font:14px/1.5 system-ui,sans-serif">Anyscale on AKS: do not hardcode a GPU SKU in Terraform <span>— Vectle</span></a>