bentoml/bentoctl-operator-template

BentoML deployment operator template

Use this template as starting point for creating your own operators to carry out deployments.

About

This is a template for a Bentoctl deployment operator.

The file structure of an operator should be similar to this. Lets look at the most important pieces.

.
├── my_operator                #1 - the operator module
│   ├── __init__.py
│   ├── create_deployable.py
│   └── generate.py
├── operator_config.py         #2 - basic config for the operator
└── README.md

#1 my_operator - This module exposes the generate(), and create_deployable() functions

Note: A good convention to use for naming operators is appending the module name with bentoctl_. This is to avoid namespace clashes with other python modules when bentoctl loads the operator. Eg bentoctl_lambda for the lambda operator module.

#2 operator_config.py - this contains the basic properties of the operator. This is accessed by the deployment tool. This container the operator name, the operator schema and the optional operator module path.

eg: lets see the sample operator_config file

OPERATOR_NAME = 'bentoml_template'

OPERATOR_MODULE_DIR= "my_operator"

OPERATOR_DEFAULT_TEMPLATE = 'terraform'

OPERATOR_AVAILABLE_TEMPLATES = ['terraform', 'kubernetes']

# BentoML deployment tool use Cerberus to validate the input. The following is an example of the schema validation.
OPERATOR_SCHEMA = {
    "project_name": {"type": "string"},
    "system_requirements": {
        "type": "dict",
        "allow_unknown": False,
        "require_all": True,
        "schema": {
            "cpu": {"type": "integer", "min": 1},
            "memory": {"type": "integer", "min": 100},
        },
    },
}

The OPERATOR_SCHEMA specfies the schema and data type for the configurable options. Bentoctl uses Cerberus for specifying and validation the schema for an operators configurable options. Reffer to the validation rules for rules that can be applyed.

Setup

This is a template repository for BentoML dpeloyment operator. To Create a new deployment operator:

1. Click "Use this template" button to start a new repository
2. Update the OPERATOR_NAME, OPERATOR_SCHEMA OPERATOR_DEFAULT_TEMPLATE in the __init__.py file