binder-control v1.1.3

binder-control

CLI for launching and managing binder server processes

binder-control is the main module for launching/managing all server-side Binder components. At it's core, it's a fairly thin wrapper around the PM2 process manager that also manages optional background services (a database, a logging stack,...) using Docker Compose.

Prerequisites

Make sure to install these services before proceeding with the installation: 1. Node v5.7.1 (easiest to install through NVM) 1. wget -qO- https://raw.githubusercontent.com/creationix/nvm/v0.31.0/install.sh | bash 2. nvm install v5.7.1 3. nvm use v5.7.1 2. PM2: npm install pm2 -g 3. Docker (optional, for background services only) 4. Docker Compose (optional, for background services only)

If you want to use the local Kubernetes VM for testing, also install: 1. Vagrant 2. VirtualBox

Installation

binder-control is designed to be installed globally and used from the command line.

npm install binder-control -g

If you are running on a GCE instance, you'll first need to download/install the gcloud utility:

curl https://sdk.cloud.google.com | bash gcloud components update

Then ensure that gcloud has the correct permissions to create new instances:

gcloud auth login

Getting Started

If you've created a new GCE instance specifically for Binder, and you'd like to proceed with the default configuration options, then binder-control start-all will interactively launch all background services and Binder servers.

To interactively stop all services/services, call binder-control stop-all

Services

binder-control provides built-in options for deploying the end-to-end system with either a local version of Kubernetes (using Vagrant with VirtualBox), or with a small, preconfigured Kubernetes cluster running on GCE.

Database

binder-control start-service db

The db service will launch a Docker container running Mongo DB (with other launch details specified in the binder-control configuration file).

Logging

binder-control start-service logging

The logging service launches three Docker containers running the ELK stack (Elasticsearch, Logstash and Kibana) for collecting and indexing log messages generated by other Binder components. If you would like to use your own logging infrastructure (i.e. use an Elasticsearch cluster for higher logging throughput), check out binder-logging for details on how to swap out the logging service with your existing infrastructure.

Kubernetes VM

binder-control start-service kube-vm

The kube-vm service, launched with the command binder-control start-service kube-vm, will spin up a virtual machine running Kubernetes, and will insert a kubectl binary onto your path that can interface with this VM. This service depends on VirtualBox and Vagrant (installation instructions here and here respectively), which must be preinstalled separately.

This Kubernetes VM will only be accessible on the local machine, and will not be accessible through the binder-web interface -- it should only be used for testing binder-deploy-kubernetes during development.

Kubernetes Cluster

binder-control start-service kube-cluster

The kube-cluster service will prompt the user for a set of cluster configuration parameters (cloud provider, cluster size, etc.) before creating a Kubernetes cluster. The cluster creation process will take ~5 minutes, after which the standard Kubernetes control commands can be issued through the kubectl.sh script, which will be downloaded the first time kube-cluster is started.

Stopping kube-cluster will permanently destroy the cluster, so make sure any important data (such as data stored in Kubernetes volumes) are backed up.

IMPORTANT: kube-cluster must be passed a desired cluster size, and creating a cluster with many nodes can get very expensive. Checkout out the GCE pricing guide to make sure your cluster size matches your budget.

Start

If you want to use the provided logging stack, DB infrastructure, or Kubernetes test VM, the very first step is to launch these services using Docker Compose:

binder-control start-service (db|logging|kubernetes)

Once the services are all up and running, you can spin up the actual Binder servers...

with a single command, using default options: binder-control start-all

or individually, with custom configuration files binder-control (build|deploy-kubernetes) start --api-key=<key> --config=/path/to/config

Manage

All services and servers launched by binder-control are managed by the PM2 process manager.

To check on the status of all binder-control managed processes, see restart counts and uptime information: pm2 list

To see any console output or logging information: pm2 logs <process name>

Stop

To stop background services, use the stop-service command binder-control stop-service (db|logging|kubernetes)

To stop Binder servers, use the server name as the subcommand binder-control (build|deploy-kubernetes) stop