portal

README

Portal is an api-driven, in-kernel layer 2/3 load balancer.

Status

Complete/Experimental

Usage:

As a CLI

Simply run portal <COMMAND>

portal or portal -h will show usage and a list of commands:

portal - load balancer/proxy

Usage:
  portal [flags]
  portal [command]

Available Commands:
  add-service    Add service
  remove-service Remove service
  show-service   Show service
  show-services  Show all services
  set-services   Set service list
  set-service    Set service
  add-server     Add server to a service
  remove-server  Remove server from a service
  show-server    Show server on a service
  show-servers   Show all servers on a service
  set-servers    Set server list on a service
  add-route      Add route
  set-routes     Set route list
  show-routes    Show all routes
  remove-route   Remove route
  add-cert       Add cert
  set-certs      Set cert list
  show-certs     Show all certs
  remove-cert    Remove cert
  add-vip        Add vip
  set-vips       Set vip list
  show-vips      Show all vips
  remove-vip     Remove vip

Flags:
  -C, --api-cert="": SSL cert for the api
  -H, --api-host="127.0.0.1": Listen address for the API
  -k, --api-key="": SSL key for the api
  -p, --api-key-password="": Password for the SSL key
  -P, --api-port="8443": Listen address for the API
  -t, --api-token="": Token for API Access
  -b, --balancer="lvs": Load balancer to use (nginx|lvs)
  -r, --cluster-connection="none://": Cluster connection string (redis://127.0.0.1:6379)
  -T, --cluster-token="": Cluster security token
  -c, --conf="": Configuration file to load
  -d, --db-connection="scribble:///var/db/portal": Database connection string
  -i, --insecure[=false]: Disable tls key checking (client) and listen on http (server)
  -j, --just-proxy[=false]: Proxy only (no tcp/udp load balancing)
  -L, --log-file="": Log file to write to
  -l, --log-level="INFO": Log level to output
  -x, --proxy-http="0.0.0.0:80": Address to listen on for proxying http
  -X, --proxy-tls="0.0.0.0:443": Address to listen on for proxying https
  -s, --server[=false]: Run in server mode
  -v, --version[=false]: Print version info and exit
  -w, --work-dir="/var/db/portal": Directory for portal to use (balancer config)

Use "portal [command] --help" for more information about a command.

For usage examples, see Api and/or Cli readme

As a Server

To start portal as a server run:

portal --server

An optional config file can also be passed on startup:

portal -c /path/to/config.json

config.json

{ "api-token": "", "api-host": "127.0.0.1", "api-port": 8443, "api-key": "", "api-cert": "", "api-key-password": "", "db-connection": "scribble:///var/db/portal", "cluster-connection": "none://", "cluster-token": "", "insecure": false, "just-proxy": false, "proxy-http": "0.0.0.0:80", "proxy-tls": "0.0.0.0:443", "balancer": "nginx", "work-dir": "/var/db/portal", "log-level": "INFO", "log-file": "", "server": true }

## API:

| Route | Description | payload | output |
| --- | --- | --- | --- |
| **Get** /services | List all services | nil | json array of service objects |
| **Post** /services | Add a service | json service object | json service object |
| **Put** /services | Reset the list of services | json array of service objects | json array of service objects |
| **Put** /services/:service_id | Reset the specified service | nil | json service object |
| **Get** /services/:service_id | Get information about a service | nil | json service object |
| **Delete** /services/:service_id | Delete a service | nil | success message or an error |
| **Get** /services/:service_id/servers | List all servers on a service | nil | json array of server objects |
| **Post** /services/:service_id/servers | Add new server to a service | json server object | json server object |
| **Put** /services/:service_id/servers | Reset the list of servers on a service | json array of server objects | json array of server objects |
| **Get** /services/:service_id/servers/:server_id | Get information about a server on a service | nil | json server object |
| **Delete** /services/:service_id/servers/:server_id | Delete a server from a service | nil | success message or an error |
| **Delete** /routes | Delete a route | subdomain, domain, and path (json or query) | success message or an error |
| **Get** /routes | List all routes | nil | json array of route objects |
| **Post** /routes | Add new route | json route object | json route object |
| **Put** /routes | Reset the list of routes | json array of route objects | json array of route objects |
| **Delete** /certs | Delete a cert | json cert object | success message or an error |
| **Get** /certs | List all certs | nil | json array of cert objects |
| **Post** /certs | Add new cert | json cert object | json cert object |
| **Put** /certs | Reset the list of certs | json array of cert objects | json array of cert objects |
| **Delete** /vips | Delete a vip | json vip object | success message or an error |
| **Get** /vips | List all vips | nil | json array of vip objects |
| **Post** /vips | Add new vip | json vip object | json vip object |
| **Put** /vips | Reset the list of vips | json array of vip objects | json array of vip objects |

- **service_id** is a formatted combination of service info: type-host-port. (tcp-127_0_0_3-80)  
- **server_id** is a formatted combination of server info: host-port. (192_0_0_3-8080)  

For examples, see [the api's readme](api/README.md)  

## Data types:
### Service:
json:
```json
{
  "host": "127.0.0.1",
  "port": 1234,
  "type": "tcp",
  "scheduler": "wlc",
  "persistence": 300,
  "netmask": "255.255.255.0",
  "servers": []
}

json with a server:

{
  "host": "127.0.0.1",
  "port": 8080,
  "type": "tcp",
  "scheduler": "wlc",
  "persistence": 300,
  "netmask": "",
  "servers": [
    {
      "host": "172.28.128.4",
      "port": 8081,
      "forwarder": "m",
      "weight": 1,
      "UpperThreshold": 0,
      "LowerThreshold": 0
    }
  ]
}

Fields:

• host: IP of the host the service is bound to.
• interface: Host interface the service is bound to (optional, overrides host).
• port: Port that the service listens to.
• type: Type of service.
  • tcp
  • udp
• scheduler: How to pick downstream server.
  • rr - Round Robin
  • wrr - Weighted Round Robin
  • lc - Least-Connection
  • wlc - Weighted Least-Connection
  • lblc - Locality-Based Least-Connection
  • lblcr - Locality-Based Least-Connection with Replication
  • dh - Destination Hashing
  • sh - Source Hashing
  • sed - Shortest Expected Delay
  • nq - Never Queue
• persistence: Timeout for keeping requests from the same client going to the same server.
• netmask: How to group clients with persistence to servers.
• servers: Array of server objects associated to the service.

Server:

json:

{
  "host": "127.0.0.1",
  "port": 1234,
  "forwarder": "m",
  "weight": 1,
  "upper_threshold": 0,
  "lower_threshold": 0
}

Fields:

• host: IP of the host the service is bound to.
• port: Port that the service listens to.
• forwarder: Method to use to forward traffic to this server.
  • g - Gatewaying
  • i - Ipip encapsulation
  • m - Masquerading
• weight: Weight to perfer this server. Set to 0 if no traffic should go to this server.
• upper_threshold: Stop sending connections to this server when this number is reached.
  • 0 - No limit
• lower_threshold: Restart sending connections when drains down to this number.
  • 0 - Not set

Vip:

json:

{
  "ip": "192.168.0.101/24",
  "interface": "eth1",
  "alias": "eth1:0"
}

Fields:

• ip: Ip to add to interface. Should be ip/cidr
• interface: Interface to add ip to
• alias: Alias to assign the ip (can be used as "interface" when adding a service)

Route:

json:

{
  "subdomain": "admin",
  "domain": "test.com",
  "path": "/admin*",
  "targets": ["http://127.0.0.1:8080/app1","http://127.0.0.2"],
  "fwdpath": "/",
  "page": ""
}

Fields:

• subdomain: Subdomain to match on
• domain: Domain to match on
• path: Path to match on
• targets: URIs of servers
• fwdpath: Path to forward to targets (combined with target path)
• page: Page to serve instead of routing to targets

Certificate:

json:

{
  "key": "-----BEGIN PRIVATE KEY-----\nMII.../J8\n-----END PRIVATE KEY-----",
  "cert": "-----BEGIN CERTIFICATE-----\nMII...aI=\n-----END CERTIFICATE-----"
}

Fields:

• key: Pem style key
• cert: Pem style certificate

Error:

json:

{
  "error": "exit status 2: unexpected argument"
}

Fields:

• error: Error message

Message:

json:

{
  "msg": "Success"
}

Fields:

• msg: Success message

Contributing

Portal uses govendor to manage dependencies. After clone, run govendor sync to ensure you have the proper dependencies prior to building.

Todo

• vip testing
• balance vips across cluster
• redis cluster init to get and set certs and vips

Changelog

• 0.1.1 - 17-06-20

  • Log and fail if db init fails
  • Use govendor for deps
  • Pre-validate protocol for services
  • Log all 'forwarded-for' ips

• 0.1.0 - 17-01-24

  • Fixed start issues related to VIPs

• Added ability to specify interface when registering a service

• Users can now add vips to portal

• Portal now uses the service host for server's hosts that are "127.0.0.1"

Documentation

Overview

Portal is an api-driven, in-kernel layer 2/3 load balancer with http/s proxy cababilities.

Usage

To run as a server, using the defaults, starting portal is as simple as

portal -s

For more specific usage information, refer to the help doc (portal -h):

Usage:
  portal [flags]
  portal [command]

Available Commands:
  add-service    Add service
  remove-service Remove service
  show-service   Show service
  show-services  Show all services
  set-services   Set service list
  set-service    Set service
  add-server     Add server to a service
  remove-server  Remove server from a service
  show-server    Show server on a service
  show-servers   Show all servers on a service
  set-servers    Set server list on a service
  add-route      Add route
  set-routes     Set route list
  show-routes    Show all routes
  remove-route   Remove route
  add-cert       Add cert
  set-certs      Set cert list
  show-certs     Show all certs
  remove-cert    Remove cert

Flags:
  -C, --api-cert="": SSL cert for the api
  -H, --api-host="127.0.0.1": Listen address for the API
  -k, --api-key="": SSL key for the api
  -p, --api-key-password="": Password for the SSL key
  -P, --api-port="8443": Listen address for the API
  -t, --api-token="": Token for API Access
  -b, --balancer="lvs": Load balancer to use (nginx|lvs)
  -r, --cluster-connection="none://": Cluster connection string (redis://127.0.0.1:6379)
  -T, --cluster-token="": Cluster security token
  -c, --conf="": Configuration file to load
  -d, --db-connection="scribble:///var/db/portal": Database connection string
  -i, --insecure[=false]: Disable tls key checking (client) and listen on http (server)
  -j, --just-proxy[=false]: Proxy only (no tcp/udp load balancing)
  -L, --log-file="": Log file to write to
  -l, --log-level="INFO": Log level to output
  -x, --proxy-http="0.0.0.0:80": Address to listen on for proxying http
  -X, --proxy-tls="0.0.0.0:443": Address to listen on for proxying https
  -s, --server[=false]: Run in server mode
  -v, --version[=false]: Print version info and exit
  -w, --work-dir="/var/db/portal": Directory for portal to use (balancer config)

Use "portal [command] --help" for more information about a command.

Build Specs

It is built with clustering at it's core to ensure syncronization between nodes. It utilizes a multi-master replication system allowing any node to accept requests to update load balancing or proxy rules. The high-level workflow is as follows:

// every call starts by hitting the api
API - setRoute
- calls cluster.SetRoute

  // in order to ensure syncronization comes first, cluster starts the work
  CLUSTER - SetRoute
  - calls publish "set-route"

    // the redis clusterer utilizes the pub/sub functionality for syncronization
    // when it recieves a message, it calls on "common" to implement the changes
    SUBSCRIBER - on "set-route"
    - calls common.SetRoute

      // common contains all the logic to perform an action, as well as roll back
      // other "systems" upon failure, effectively "undoing" the action
      COMMON - SetRoute
      - calls proxymgr.SetRoute & database.SetRoute
      - rolls back proxymgr if database fails

    SUBSCRIBER
    - if common.SetRoute was successful, write success to redis for self, otherwise
      rollback self

  // the cluster member that received the request ensures all members got the update
  CLUSTER
  - returns err if not all members have set route
  - rolls back `common` (via publish) if not all members can set route

API
- if error, return 500 response, otherwise respond as fits

Portal is also extremely "pluggable", meaning that it is easy to code in another tool to be a "Balancer" or "Proxy" by matching the interfaces. Individual object (database|balancer|proxy) functions are never called directly by "common". "Common" calls package level functions such as

database.SetService

which calls/sreturns the selected "Backend's" SetService function. The type of backend (scribble/redis) is determined from the connection configuration option, as seen in the package's Init() function:

url, err := url.Parse(config.ClusterConnection)
if err != nil {
	return fmt.Errorf("Failed to parse db connection - %s", err)
}

switch url.Scheme {
case "redis":
	Clusterer = &Redis{}
case "none":
	Clusterer = &None{}
default:
	Clusterer = &None{}
}

return Clusterer.Init()