RUNNING.md

Running the O² control system

⚠️ 🚧 this section is out of date

User? See Single node O²/FLP software deployment instructions

Developer? See the run script and the configuration files in the hacking directory instead

This part assumes you have already set up the Go environment, fetched the sources and built o2control-core and o2control-executor in bin, deployed the DC/OS Vagrant development environment and set up the software you wish to run on this cluster (for example with fpctl).

grpcc

In order to talk to o2control-core we can use coconut, or we can make calls directly with a gRPC client, such as grpcc.

Assuming you have installed Node.js and npm (on CC7 $ sudo yum install http-parser nodejs npm), the installation with npm is straightforward.

$ sudo npm install -g grpcc

Putting it all together

Assuming the DC/OS Vagrant environment is up, a Mesos master will be running at m1.dcos with:

• DC/OS interface at http://m1.dcos/,
• Mesos interface at http://m1.dcos/mesos/,
• Marathon interface at http://m1.dcos/marathon/.

The hacking directory contains some wrapper scripts that rely on a Mesos master at m1.dcos and make running o2control-core easy.

It also contains a dummy configuration file (config.yaml) which simulates what should normally be a Consul instance.

Run o2control-core:

$ hacking/run.sh

or:

$ bin/o2control-core -mesos.url http://m1.dcos:5050/api/v1/scheduler -executor.binary </in-cluster/path/to/o2control-executor> -verbose -config "file://hacking/example-config.yaml"

Use grpcc to talk to it:

$ hacking/grpcc.sh

or:

$ grpcc -i --proto core/protos/o2control.proto --address 127.0.0.1:47102

See Using coconut for instructions on the O² Control core command line interface.