Docker Ubuntu upstart scripts

README.md

Production Ready Process Monitoring

1. Install docker-manger.conf and docker-instance.conf in /etc/init

2. Create a containers file in /etc/docker/ with the following format

    name: image cmd
   

   The name must be unique.

3. sudo service docker-manager start

4. You should now have upstart monitoring your container.

5. Use docker stop to kill your container and notice it gets restarted

Due to a bug in upstart (https://bugs.launchpad.net/upstart/+bug/568288) doing service docker-instance stop NAME=XXXX fails to stop the instance. I'm not actually sure how to work around this.

containers

nginx-0: docker.wdig.com/cws/base-service nginx
nginx-1: docker.wdig.com/cws/base-service nginx
nginx-2: docker.wdig.com/cws/base-service nginx

docker-instance.conf

description     "Docker Container"

stop on stopping docker
# Docker has a timeout of 10 seconds so as long as this
# is longer so we don't kill the wait process
kill timeout 20

# We don't want to TERM the `docker wait` process so we fake the signal 
# we send to it. The pre-stop script issues the `docker stop` command
# which causes the `docker wait` process to exit
kill signal CONT

# Due to a bug in upstart we need to set the modes we consider
# successful exists https://bugs.launchpad.net/upstart/+bug/568288
normal exit 0 CONT

respawn

instance ${NAME}

pre-start script
  [ ! -f /etc/docker/containers ] && { stop; exit 0; }

  if ! grep -q "^${NAME}:" /etc/docker/containers; then
    stop
    exit 0
  fi
end script

script
  LINE=$(grep "^${NAME}:" /etc/docker/containers | cut -d: -f2-| sed 's/^\s*//')

  IMAGE=$(echo "$LINE" | cut -d\  -f1)
  CMD=$(echo "$LINE" | cut -d\  -f2- -s)
  
  [ -f "/var/run/docker/${NAME}.cid" ] && { ID="$(cat /var/run/docker/${NAME}.cid)"; }
  
  START=1
  if [ ! -z "${ID}" ]; then
    if docker ps | grep -q "${ID}"; then
      START=0
    else
      NID=$(docker start ${ID})
      [ "${NID}" = "${ID}" ] && START=0
    fi
  fi

  if [ $START -ne 0 ]; then
    ID=$(docker run -d "${IMAGE}" ${CMD})
    echo ${ID} > /var/run/docker/${NAME}.cid
  fi

  exec docker wait ${ID}
end script

pre-stop script
  [ -f "/var/run/docker/${NAME}.cid" ] && { ID="$(cat /var/run/docker/${NAME}.cid)"; }
  if [ ! -z "${ID}" ]; then
     docker stop "${ID}"
  fi
end script

docker-manager.conf

description     "Manage Docker Containers"

start on started docker
stop on stopping docker

respawn

pre-start script
  # Directory to store run data
  [ ! -d /var/run/docker ] && mkdir -p /var/run/docker

  # Directory to store config
  [ ! -d /etc/docker ] && mkdir -p /etc/docker

  [ ! -f /etc/docker/containers ] && { stop; exit 0; }

  while read line; do
    NAME=$(echo "$line" | cut -d: -f 1)
    start docker-instance NAME="$NAME"
  done < /etc/docker/containers
end script

post-stop script
    for inst in `initctl list | grep "^docker-instance "|awk '{print $2}'|tr -d ')'|tr -d '('`
    do
        stop docker-instance NAME=$inst
    done
end script

To get this to work with images that are built with a CMD default, I replaced the docker run in line 49 with:

    if [ "${IMAGE}" = "${CMD}" ]; then
      ID=$(docker run -d "${IMAGE}")
    else
      ID=$(docker run -d "${IMAGE}" ${CMD})
    fi

The script should be fixed to start with the default command now.

Thanks for the starting point. I had two problems with your script. 1) as written, it prefers to run old containers rather than pulling new ones when appropriate. Deleting the 'start' clause fixes that. 2) One of my major reasons for using upstart is because it does logs better than docker. So I changed your docker run -d into an exec docker run, and then used ps -l -q in a post-start script to create the run file.