NOTE

I haven't maintained this for a couple of years now, Ansible has moved on a bit.

Also the base box will need an update.

Mesos cluster

CentOS 7 based Mesos cluster. Intended to be a clean baseline for framework testing.

Currently 0.28.1 with docker support and cgroups isolation (see roles/mesos/{slave,master}/defaults for various tunings).

See frameworks/ for some ways to run frameworks against this stack.

base OS

CentOS 7.x x64 (from https://github.com/box-cutter/centos-vm ) - see Vagrantfile for precise version.

requirements

the centos base image above
Ansible (2.x) on your host machine
Virtualbox on host machine (box has 4.3.28 extensions)
Vagrant 1.8.x or better (for linked_clones)
2 Vagrant plugins (see below)

assumptions

Hosts permit passwordless ssh+sudo for the relevant ansible_ssh_user.
hosts connect to each others inventory hostname
DNS works (see previous point)

Vagrant setup

an inventory for Vagrant is in vagrant/hosts, the hostnames in there need to match your Vagrantfile.

(NB: you won't get a lot of RAM offered by the slaves. Mesos > 0.20.0 reserves 1Gb or half system RAM for the OS)

name resolution

we'll use a Vagrant plugin to ensure all nodes (and our host) can resolve each others names.

The hostmanager plugin will auto-manage that.

vagrant plugin install vagrant-hostmanager

dedicated storage for /var/mesos on the slaves

This should stop slaves being taken out when sandboxes overflow, and also allow some custom mount options (see Vagrantfile for detail). the storage attaches at /dev/sdb

vagrant plugin install vagrant-persistent-storage

Disks live under disks/ at the top of this repo.

do the thing

vagrant up

NB: hostmanager will (try to) edit your local /etc/hosts

wiping

If you want to blow everything away, this should do the trick:

vagrant destroy -f
for i in master1 slave1 slave2
  do
    ssh-keygen -R $i
  done
# remove the disks via Virtualbox 'Virtual Media Manager'

time for Ansible

run the main play with:

ansible-playbook -i vagrant/ site.yml

If you add/destroy vagrant VMs, the 'vagrant up' should auto-manage your local /etc/hosts along with existing VMs. If you need to ensure it's up to date, just run

vagrant hostmanager

vars

role-specific vars live in $rolename/defaults/main.yml
'environment' specific vars (e.g. ansible_ssh_user) live in [all:vars] within the relevant inventory and override role defaults

BUGS

Oh, I expect so. Log an issue / PR if you notice any.

rasputnik / mesos-centos