wangxin688/netpalm

ReST based network device broker

The Open API Platform for Network Devices

netpalm makes it easy to push and pull state from your apps to your network by providing multiple southbound drivers, abstraction methods and modern northbound interfaces such as open API3 and REST webhooks.

Supporting netpalm

Apcela Because Enterprise Speed Matters	Delivering the power to communicate	Maybe you?

Table of Contents

• What is netpalm?
• Features
• Concepts
• Additional Features
• Examples
• API Docs
• Caching
• Configuration
• Installation
• Further Reading
• Contributing

What is netpalm?

Leveraging best of breed open source network components like napalm, netmiko, ncclient and requests, netpalm makes it easy to abstract from any network devices native telnet, SSH, NETCONF or RESTCONF interface into a modern model driven open api 3 interface.

Taking a platform based approach means netpalm allows you to bring your own jinja2 config, service and webhook templates, python scripts and webhooks for quick adoption into your existing devops workflows.

Built on a scalable microservice based architecture netpalm provides unparalleled scalable API access into your network.

Features

• Speaks REST and JSON RPC northbound, then CLI over SSH or Telnet or NETCONF/RESTCONF southbound to your network devices
• Turns any Python script into a easy to consume, asynchronous and documented API with webhook support
• Large amount of supported network device vendors thanks to napalm, netmiko, ncclient and requests
• Built in multi-level abstraction interface for network service lifecycle functions for create, retrieve and delete and validate
• In band service inventory
• Ability to write your own service models and templates using your own existing jinja2 templates
• Well documented API with postman collection full of examples and every instance gets it own self documenting openAPI 3 UI.
• Supports pre- and post-checks across CLI devices raising exceptions and not deploying config as required
• Multiple ways to queue jobs to devices, either pinned strict (prevent connection pooling at device)or pooled first in first out
• Modern, container based scale out architecture supported by every component
• Highly configurable for all aspects of the platform
• Leverages an encrypted Redis layer providing caching and queueing of jobs to and from devices

Concepts

Basic Concepts

netpalm acts as a ReST broker and abstraction layer for NAPALM, Netmiko, NCCLIENT or a Python Script. netpalm uses TextFSM or Jinja2 to model and transform both ingress and egress data if required.

Component Concepts

netpalm is underpinned by a container based scale out architecture for all components.

Queueing Concepts

netpalm provides domain focused queueing strategy for task execution on network equipment.

Scaling Concepts

Every netpalm container can be scaled in and out as required. Kubernetes or Swarm is recommended for any large scale deployments.

To scale out the basic included compose deployment use the docker-compose command

docker-compose scale netpalm-controller=1 netpalm-worker-pinned=2 netpalm-worker-fifo=3

Additional Features

• Jinja2

  • BYO jinja2 config templates
  • BYO jinja2 service templates
  • BYO jinja2 webhook templates
  • Can be used to just render Jinja2 templates via the REST API
  • Automatically generates a JSON schema for any Jinja2 Template

• Parsers

  • TextFSM support via netmiko
  • NTC-templates for parsing/structuring device data (includes)
  • TTP Template Text Parser - Jinja2-like parsing of semi-structured CLI data
  • Napalm getters
  • Genie support via netmiko
  • Automated download and installation of TextFSM templates from http://textfsm.nornir.tech online TextFSM development tool
  • Optional dynamic rendering of Netconf XML data into JSON

• Webhooks

  • Comes with standard REST webhook which supports data transformation via your own jinja2 template
  • Supports you to bring your own (BYO) webhook scripts

• Scripts

  • Execute ANY python script as async via the ReST API and includes passing in of parameters
  • Supports pydantic models for data validation and documentation

• Queueing

  • Supports a "pinned" queueing strategy where a dedicated process and queue is established for your device, tasks are sync queued and processed for that device
  • Supports a "fifo" pooled queueing strategy where a pool of workers
  • Supports on the fly changes to the async queue strategy for a device

• Caching

  • Can cache responses from devices so that the same request doesn't have to go back to the device
  • Automated cache poisioning on config changes on devices

• Scaling

  • Horizontal container based scale out architecture supported by each component

Examples

We could show you examples for days, but we recommend playing with the online postman collection to get a feel for what can be done. We also host a public instance where you can test netpalm via the Swagger UI.

getconfig method

netpalm also supports all arguments for the transport libs, simply pass them in as below

check response

ServiceTemplates

netpalm supports model driven service templates, these self render an OpenAPI 3 interface and provide abstraction and orchestration of tasks across many devices using the get/setconfig or script methods.

The below example demonstrates basic SNMP state orchestration across multiple devices for create, retrieve, delete

Template Development and Deployment

netpalm is integrated into http://textfsm.nornir.tech so you can ingest your templates with ease

API Docs

netpalm comes with a Postman Collection and an OpenAPI based API with a SwaggerUI located at http://localhost:9000/ after starting the container.

Caching

• Supports the following per-request configuration (/getconfig routes only for now)

  • permit the result of this request to be cached (default: false), and permit this request to return cached data
  • hold the cache for 30 seconds (default: 300. Should not be set above redis_task_result_ttl which defaults to 500)
  • do NOT invalidate any existing cache for this request (default: false)

    {
      "cache": {
        "enabled": true,
        "ttl": 30,
        "poison": false
      }
    }

• Supports the following global configuration:

  • Enable/Disable caching: "redis_cache_enabled": true for caching to apply it must be enabled BOTH globally and in the request itself
  • Default TTL: "redis_cache_default_timeout": 300

• Any change to the request payload will result in a new cache key EXCEPT:

  • JSON formatting. { "x": 1, "y": 2 } == {"x":1,"y":2}
  • Dictionary ordering: {"x":1,"y":2} == {"y":2,"x"1}
  • changes to cache configuration (e.g. changing the TTL, etc)
  • fifo vs pinned queueing strategy

• Any call to any /setconfig route for a given host:port will poison ALL cache entries for that host:port

  • Except /setconfig/dry-run of course

Configuration

Edit the config/config.json file to change any parameters ( see defaults.json for example )

Installation

1. Ensure you first have docker installed

sudo apt-get install docker.io
sudo apt-get install docker-compose

2. Clone this repository

git clone https://github.com/tbotnz/netpalm.git
cd netpalm

3. Build the container

sudo docker-compose up -d --build

4. After the container has been built and started, you're good to go! netpalm will be available on port 9000 under your docker hosts IP.

http://$(yourdockerhost):9000

Further Reading

• Cisco Developer Portal

• Wim Wauters - netpalm Intro Part 1

• Wim Wauters - netpalm Intro Part 2

• Wim Wauters - netpalm Intro Part 3

• NetworkCollective w/ Jason Edelman - Podcast Episode about NTC / netpalm

• Packetflow - Top 5 Up and Coming Network Automation Tools

• ipspace - Building Multivendor Network Automation Platform

• ipspace - Useful Network Automation Tools

Contributing

We are open to contributions, before making a PR, please make sure you've read our CONTRIBUTING.md document.

You can also find us in the channel #netpalm on the networktocode Slack.