Automation-ai pod fails to start with "FailedMount" error when using Ollama.

Products

• CA Automic Workload Automation - Automation Engine
• Automic SaaS

Issue/Introduction

When attempting to implement Ollama AI instead of Google Vertex AI (Gemini) in a Kubernetes environment (AAKE), the automation-ai pod fails to start.

Checking the Kubernetes events (kubectl describe pod) reveals the following error:

Warning  FailedMount  39s (x9 over 2m46s)  kubelet  MountVolume.SetUp failed for volume "vertex-ai-gemini-credentials" : references non-existent secret key: SPRING_AI_VERTEX_AI_GEMINI_CREDENTIALS

The pod remains in a ContainerCreating or Pending state and never reaches Running.

Environment

Product: Automic Automation 

Component: Automation.AI 

Version Affected: 24.4.x

Cause

This issue is caused by a defect in the deployment configuration of the automation-ai component (tracked as DE184303).

The deployment template contains a mandatory volume mount for the Google Vertex AI credentials secret (vertex-ai-gemini-credentials). Kubernetes prevents the pod from starting if this secret or the specific key SPRING_AI_VERTEX_AI_GEMINI_CREDENTIALS is missing, even if the application is correctly configured to use Ollama via AUTOMATION_AI_MODEL_NAME: ollama in the ConfigMap.

Resolution

This issue is currently under investigation by Engineering under defect DE184303. A permanent fix is planned for a future release which will make this volume mount conditional based on the selected AI model.

Component: Automation AI

Automation AI 24.0.6 - TBA

Additional Information

• Ensure that AUTOMATION_AI_MODEL_NAME: ollama is used in the automation-ai-configmap.
• Related Article: 429993 - Error: Default credentials not found when configuring Ollama AI (discusses model keyword configuration).