[BUG] Unable to redeploy a model once undeployed on Windows

[rbhavna]

What is the bug? ML Commons: We have been seeing a bug on deploy/undeploy model due to Windows holding folder access while OpenSearch process is running.

• This is only a Windows isolated issue.
• It will prevent cleanup during undeploy and re-deployment on the same id.
• The folder will be released only when you stop the opensearch cluster.

How can one reproduce the bug? Steps to reproduce the behavior: 1) Register a model version

POST /_plugins/_ml/models/_register
{
  "name": "huggingface/sentence-transformers/all-MiniLM-L12-v2",
  "version": "1.0.1",
  "model_format": "TORCH_SCRIPT"
}

## Response
{
  "task_id": "FFe1cIkBT5fLQDwZTj7J",
  "status": "CREATED"
}

2) Get model_id using the task_id above:

GET /_plugins/_ml/tasks/<task_id>

3) Deploy using the model_id

POST /_plugins/_ml/models/<model_id>/_deploy

4) Undeploy it successfully

POST /_plugins/_ml/models/<model_id>/_undeploy

5) Redeploy the same model using the same model ID. The model fails to get deployed

POST /_plugins/_ml/models/<model_id>/_deploy

What is the expected behavior? When a user undeploy a model and tries to re-deploy the same model, it should get deployed successfully.

What is your host/environment?

• OS: Windows

Do you have any screenshots? If applicable, add screenshots to help explain your problem.

Do you have any additional context? Add any other context about the problem.