Proof of concept and playground for a daemontools written in go. The goal is to have a stable service starter which automagically restarts a service if it has been shut down and log (with logrotation) all output.

You may start up with an empty (or not even existing) service-dir and create the services after starting go-supervise. As soon as a "run" executable is found, the service is being started.

You may delete a service sub-directory (e.g. ~/services/srv1) while go-supervise is running. It will detect the deletion and shut down the service.

For now, go-supervise uses etcd to save known services, because later on

• It should be possible to add a service in etcd database and let go-supervise starts it up on the correct server.
• It should be possible to remove a service in etcd database and let go-supervise shut it down.

etcd is expected on 127.0.0.1:2379, which is default. Later on go-supervise will have a configuration file.

Not existent yet, Sorry :-)

Antigloss' Logger CoreOS etcd client Vektra's TAI64n

(Please make sure, GOPATH is set and etcd is running)

mkdir -p ~/services/srv1 && echo "echo \"starting up\" ; sleep 1000" > ~/services/srv1/run && chmod 755 ~/services/srv1/run
go get github.com/adar/go-supervise
# it will claim that there are no buildable go-files - ignore

# change service-path (edit config.json - ServiceConfig/path
cd $GOPATH/src/github.com/adar/go-supervise/config
vi config.json

cd ../svscan && ./run.sh & tail -f /var/log/syslog 
# where syslog might be /var/log/messages in CentOS

(If you see "Error while getting DB client client: etcd cluster is unavailable or misconfigured" in your syslog, etcd is not running correctly.)

See issues page

MIT license, see LICENSE.txt for details.