cvproxy 1.0.3

CVProxy - CloudVision Proxy

Arista CloudVision Proxy

Arista are transitioning their CloudVision provisioning model to use Studios and Workspaces, but the HTTP REST API only supports the legacy provisioning model, which has now been deprecated. The new API uses gRPC and is rather complex, but Arista provides a Python library to talk to the API, which is used by Ansible via the arista.avd.cv_deploy role. The purpose of this tool is to use pyavd to handle the complexity and to provide a HTTP proxy to convert between simplified JSON and gRPC using the same workflow that Ansible uses via pyavd._cv.workflows.deploy_to_cv.

Installation

python3 -m venv /opt/cvproxy
/opt/cvproxy/bin/python3 -m pip install cvproxy

tee /etc/systemd/system/cvproxy.service >/dev/null <<-EOF
[Unit]
Description=CVProxy - CloudVision Proxy

[Service]
Environment="VIRTUAL_ENV=/opt/cvproxy"
ExecStart=/opt/cvproxy/bin/python3 -u -m cvproxy -s -l 127.0.0.1 -p 8080
SyslogIdentifier=cvproxy
TimeoutStartSec=60
Restart=always

[Install]
WantedBy=multi-user.target
EOF

systemctl enable --now cvproxy

Usage

 cvproxy -s [-l <address>] [-p <port>] [-xff]

   -s                         - start CVProxy
   -l <address>               - specify a listen address (default is '127.0.0.1')
   -p <port>                  - specify a listen port (default is 8080)
   -xff                       - use X-Forwarded-For

It works by accepting a HTTP POST request with a Content-Type of application/json, which should adhere to the following schema:

"unevaluatedProperties": false,
"required": ["devices", "cv_server", "cv_token"],
"properties": {
  "devices": {
    "minProperties": 1,
    "unevaluatedProperties": false,
    "patternProperties": {
      "^[a-z][a-z0-9_.-]*$": {
        "unevaluatedProperties": false,
        "required": ["configlet"],
        "properties": {
          "serial_number": { "type": "string", "pattern": "^[A-Z][A-Z0-9]{10}$" },
          "configlet": { "type": "string", "pattern": "^(?=(.{4})+$)[A-Za-z0-9+/-]+={0,2}$" },
          "tags": {
            "minProperties": 1,
            "additionalProperties": { "type": "string", "pattern": "\\S+" }
          }
        }
      }
    }
  },
  "cv_server": { "type": "string", "pattern": "\\S+" },
  "cv_token": { "type": "string", "pattern": "\\S+" },
  "cv_change_control_name": { "type": "string", "pattern": "\\S+" },
  "cv_delete_workspace": { "type": "boolean" },
  "cv_strict_tags": { "type": "boolean" }
}

The serial_number attribute is optional as it should be able to identify the device based on the hostname, unless it is a new device and there is no existing configlet. The configlet attribute is mandatory and is the configlet for the device, which must be Base64 encoded.

Example:

"devices": {
  "leaf-1a": {
    "configlet": "<base64 encoded configlet>",
    "tags": {
      "type": "leaf"
    }
  },
  "leaf-1b": {
    "configlet": "<base64 encoded configlet>",
    "tags": {
      "type": "leaf"
    }
  }
},
"cv_server": "www.cv-prod-uk-1.arista.io",
"cv_token": "<service token>",
"cv_strict_tags": true

If the HTTP POST request succeeds and there are any changes then a Change Control will be created in CloudVision ready for you to review and approve. The cv_change_control_name attribute is optional and if not provided will use the AVD default. The cv_delete_workspace attribute is also optional and if set to "true" will cleanup and delete the submitted Workspace. The cv_strict_tags attribute is also optional and defaults to "false".

Successful HTTP responses will be JSON encoded and will always contain a status attribute, which will either be "ok" or "error". If it is set to "ok" then an optional change_control attribute will be included if a Change Control was generated. If it is set to "error" then an errors attribute will also be provided with details of why it failed.

The recommended deployment is to deploy HAProxy in front of CVProxy, as HAProxy is better equipped to deal with TLS termination and it also supports HTTP/2 and HTTP/3 transports.