flynn-host is the Flynn host service. An instance of it runs on every host in
the Flynn cluster. It is responsible for running jobs (in Linux containers) and
reporting back to schedulers and the leader.
flynn-host is capable of bootstrapping itself inside of containers with its dependencies using the included manifest. It depends on discoverd, which in turn depends on etcd. These three components are referred to as "Layer 0".
One instance of host acts as the "cluster leader" and serializes the creation of new jobs. The leader is maintains the entire cluster state (a list of hosts and their running jobs) in memory. If the leader disappears, a new one is elected by discoverd and the rest of the hosts connect to it and provide their current state.
The host service expects to be run inside of a Docker container with access to the Docker socket. It also needs to know the external IP address that can be used to reach it (and every container it runs inside of Docker).
IP=$(ifconfig eth0 | grep 'inet addr:' | cut -d: -f2 | awk '{print $1}')
docker run -v=/var/run/docker.sock:/var/run/docker.sock -p=1113:1113 flynn/host -external $IP
Subsequent hosts should pass a list of comma separated peers to etcd containing at least one running host:
docker run -v=/var/run/docker.sock:/var/run/docker.sock -e=ETCD_PEERS=10.1.2.10:7001 -p=1113:1113 flynn/host -external $IP
The etcd discovery
service
may also be used by setting the ETCD_DISCOVERY environment variable.
The -force option, if provided, will terminate all existing containers that
were booted by a previous instance of the host service.
- Increase test coverage
- Documentation