Scruffy

© Mütli Corp. By Dave Newman.

1. collect the state of the world

• collect AWS state (query AWS)
• collect pinky state (read from redis)
• collect prism state (read from redis)

2. look for problems

• duplicate servers running
• aws box appeared
• aws box disappeared
• server appeared
• server disappeared

3. rebalance (potentially move to brain)

• shutdown empty Shared servers
• bring up new boxes when at capacity
• shut down boxes when excess capacity
• rebalance servers for player counts

4. save new state

   /pinky/* /state /pinky/* /servers/state /prism//players/ /server//state /server//allocation (ram/ecus)

   /scruffy/cache/boxes

   /scruffy/cache/servers /scruffy/cache/pinkies /scruffy/cache/prisms

One box has one pinky, both identified by the instance-id from Amazon (eg i-fb0c3287)

Box - Amazon ec2 instance id started_at instance_type state (starting|up|stopping) # this is the state of the Amazon instance

Pinky id # same as Box#id state (starting|up|stopping) servers players freeDiskMb freeRamMb idleCpu

Server

• id
• state (importing|starting|up|stopping|exporting)

Player

• id

pinky coming up

• (scruffy creates EC2 instance - gets instance id)
• (scruffy sets box to starting)
• box - starting
• box - up
• (scruffy sets pinky state to starting)
• pinky - starting
• pinky - up

pinky going down

• (scruffy sets pinky state to stopping)
• (scruffy sets box state to stopping)
• (scruffy terminates box)

Usage

foreman start

Environment

Required

AWS_ACCESS_KEY
AWS_SECRET_KEY
AWS_REGION
REDIS_URL
BUGSNAG

Optional

SCRUFFY_ENV (staging|production)
SCRUFFY_ROOT