Standard_ND96is_flex_H100_v5 Is not supported due to not being in the whitelist for H100 SKUs
- Dominant language
- Go
- Stars
- 159
- Forks
- 277
- Avg merge
- 1d 16h
- Merged PRs (30d)
- 178
Description
**What happened**: Deploy Standard_ND96is_flex_H100_v5 userNodes. Fabric manager fails to get installed making the GPUs not usable.
**What you expected to happen**: For the nodes to come up like other H100 SKUs
**How to reproduce it**: deploy SKU Standard_ND96is_flex_H100_v5 in AKS
**Anything else we need to know?**:
**Environment**:
- AgentBaker version:
- Kubernetes version (use `kubectl version`):
- OS (e.g. from /etc/os-release):
- Kernel (e.g. `uname -a`):
- Install tools:
- Others:
Contributor guide
Research direction
Search the AgentBaker codebase for the H100 SKU whitelist and the provisioning path used when deploying Standard_ND96is_flex_H100_v5 in AKS. Run the focused tests for that path, and consider the issue done when the SKU is accepted and Fabric Manager installs so the GPUs are usable.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- azure, go, kubernetes
- Domain
- cloud, infrastructure
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 68/100