goodwritehq/compose-mesos

Run Mesos in Docker with Docker Compose!

Run Mesos in Docker with Docker Compose!

Assuming you've already installed Docker Toolbox, then checkout and run:

docker-compose up -d

To stream log output, simply run:

docker-compose logs

Great! Now let's pop open the Mesos, Marathon, & Chronos' UI like so:

open http://localdocker:5050/ # open Mesos UI
open http://localdocker:8080/ # open Marathon UI
open http://localdocker:4400/ # Open Chronos UI

Let's launch a simple app through Marathon's REST API like so:

curl -X POST -H "Accept: application/json" -H "Content-Type: application/json" \
  localdocker:8080/v2/apps -d '
{
    "id": "hello",
    "cpus": 0.1,
    "mem": 32,
    "cmd": "echo hello; sleep 10"
}'

And now for a Docker container:

curl -X POST -H "Accept: application/json" -H "Content-Type: application/json" \
  localdocker:8080/v2/apps -d '
{
    "id": "inky",
    "container": {
        "docker": {
            "image": "mesosphere/inky"
        },
        "type": "DOCKER",
        "volumes": []
    },
    "args": ["hello"],
    "cpus": 0.2,
    "mem": 32.0,
    "instances": 1
}'

Much thanks to the following sources:

• An Introduction to Mesosphere
• How To Configure a Production-Ready Mesosphere Cluster
• Marathon v0.7.0 — Running Dockers at Scale and More

Cheers!

compose-mesos's Issues

Use goforever instead of supervisor

Supervisor is terrible.

RUN apt-get -y install golang
RUN mkdir /go
RUN export GOPATH=/go && export PATH=/go/bin:$PATH
RUN go get github.com/gwoo/goforever
ADD mesos-slave.tml /mesos-slave.tml

Unable to follow "Sandbox" links from Mesos UI

After running the container example from the README and then going to the Mesos UI, this error is received when clicking on the "Sandbox" link for a running task:

Failed to connect to slave '20140927-181543-486543788-5050-1-0' on '7af7848f5538:5051'.
Potential reasons:
The slave's hostname, '7af7848f5538', is not accessible from your network
The slave's port, '5051', is not accessible from your network
The slave timed out or went offline

Screenshot:

Create fig feature request for variable sequences like in factory girl

Example from README spawns hello app that waits forever

This is likely a simple oversight in setup, but I have Docker Toolbox set up on Mac OSX (El Capitan) and docker-compose up works fine. If I edit /etc/hosts to have "localdocker" mapped to my docker-machine VM, then I can see the UIs. I can curl to localdocker and see the "hello" app land. But it stays in "waiting" forever. Should there be a "localdocker" setting in /etc/hosts in the docker containers? Is this an artifact of docker-machine being different than the old docker-osx, then boot2docker? What am I missing?

Lock versions

Right now none of the containers have versions locked - this means that potentially this combination of containers could drift and become broken. Let's lock those down so we can actually test upgrades by hand :)

Update marathon to configure with environmental variables.

Looks like they finally got it merged:

mesosphere/marathon#475

Switch container example to long running process.

Perhaps coreos/apache will work. Find something small with a helloworld.

Use progrium/busybox-docker

Use progrium/busybox-docker instead of dind for the docker-mesos container. Use progrium/busybox for chronos, marathon, etc...