scalepot

You can scale your ec2 instances automatically using scalepot.

Getting Started

• Install scalepot.

    $ git clone git@github.com:limsangjin12/scalepot.git && cd scalepot
    $ python setup.py install
  

  or

    $ pip install scalepot
  

• Write scalepot.py.

    from scalepot.api import scale, State, g as scaleinfo        
  
    def deploy_worker(ip):
        # deploy worker here
        pass
  
    @scale
    def scale():
        if scaleinfo.state == State.SCALE_OUT:
            if scaleinfo.role.name == 'worker':
                ip = scaleinfo.role.instance.ip_address
                deploy_worker(ip)
  

• Write scalepot.yml in the same path.

    az: ap-northeast-1
    roles:
    - name: worker
      type: c1.medium
      min: 2
      max: 5
      ami: ami-86db6187
      option: spot # on-demand or spot
      cooltime: 20 # minutes between checking
                   # each instances
      placement: b
  

• You need ~/.boto file also.

    [Credentials]
    aws_access_key_id = (your access key id)
    aws_secret_access_key = (your secret access key)
  

• Then,

    $ scalepot
  

Define custom checking function

• Define a function with check decorator.

    from scalepot.api import scale, State, g as scaleinfo
    from scalepot.api import check
  
    def deploy_worker(ip):
        # deploy worker here
        pass
  
    @scale
    def scale():
        if scaleinfo.state == State.SCALE_OUT:
            if scaleinfo.role.name == 'worker':
                ip = ip = scaleinfo.role.instance.ip_address
                deploy_worker(ip)
  
    @check
    def check():
        if scaleinfo.count < 3:
            return State.SCALE_OUT
        return State.NORMAL
  

Tips

• scalepot basically uses CPUUtilization metric for last 10 minutes as a default to scale-out/down.

• You can define scale_out_threshold (default value is 60) and scale_down_ratio (default value is 0.7) on your scalepot.yml file.

• scalepot.api.g.instance is a boto.ec2.instance.Instance object.

• scalepot.api.State is implemented like below.

    class State(object):
        SCALE_OUT = 'SHOULD SCALE OUT'
        SCALE_DOWN = 'SHOULD SCALE DOWN'
        MAX_LIMIT = 'MAX LIMIT REACHED'
        MIN_LIMIT = 'MIN LIMIT REACHED'
        NORMAL = 'NORMAL'
  

• To show more console options,

    $ scalepot -h