CMK cluster-autoscaler Pod in CrashLoopBackOff With "Failed to construct template node info for node group" Error

Overview

Crusoe Managed Kubernetes cluster-autoscaler is in CrashLoopBackOff status with the following error logs:

crusoe-system   cluster-autoscaler-****-****   0/1   CrashLoopBackOff   3   10.234.***.***   np-****-7.<region>.compute

E1016 12:49:12.410610   1   crusoe_node_group.go:355] Failed to construct template node info for node group <node-group-id>: instance type h100-80gb-sxm.8x
panic: runtime error: invalid memory address or nil pointer dereference [signal SIGSEGV: segmentation violation code=0x1 addr=0x180 pc=0x1776c17]

Prerequisites

• Access to the Kubernetes cluster via kubectl.

• The Crusoe CLI (crusoe) installed and authenticated.

Resolution

The cause of the issue is an invalid instance type selected for the CMK's NodePool.

In the above example, h100-80gb-sxm.8x was selected. This is not a valid instance type. It should be h100-80gb-sxm-ib.8x.

To check the available instance types:

# crusoe compute vms types

To resolve the issue, create a NodePool with a valid instance type.