Spin up your Mesos cluster with Vagrant! (Both Virtualbox and AWS are supported.)
This spins up Mesos 0.22.1 cluster and also spins up a framework server node in which Marathon (0.8.2) and Chronos (2.1.0) are running. This means you can build your own Mesos+Marathon+Chronos+Docker PaaS with vagrant up
!! Marathon works as distributed init.d
and Chronos works as distributed cron
!! If you wanted to deploy docker containers, please refer to the chapter "Deploy Docker Container with Marathon" in this blog entry.
The mesos installation is powered by Mesos chef cookbook. Please see everpeace/cookbook-mesos.
Base boxes used in Vagrantfile
s are Mesos pre-installed boxes, everpeace/mesos shared on Vagrant Cloud.
$ vagrant plugin install vagrant-omnibus
$ vagrant plugin install vagrant-berkshelf
$ vagrant plugin install vagrant-hosts
$ vagrant plugin install vagrant-cachier
$ vagrant plugin install vagrant-aws
It's so simple!
$ git clone https://github.com/everpeace/vagrant-mesos.git
$ cd vagrant-mesos/standalone
$ vagrant up
After box is up, you can see services running at:
Set ec2 credentials and some configurations defined in standalone/aws_config.yml
. You have to fill up EDIT_HERE
parts. Security group you'll set must accept at least tcp port 22(SSH) and 5050(mesos-master web ui) from outside of ec2.
# Please set AWS credentials
access_key_id: EDIT_HERE
secret_access_key: EDIT_HERE
# Please choose one from
# ["ap-northeast-1", "ap-southeast-1", "eu-west-1", "sa-east-1", "us-east-1",
# "us-west-1", "ap-southeast-2", "us-west-2"]
region: us-east-1
# array of security groups. e.g. ['sg*** ']
security_groups: EDIT_HERE
# See http://aws.amazon.com/ec2/instance-types/#selecting-instance-types
# for other instance types and its specs.
instance_type: m1.small
keypair_name: EDIT_HERE
ssh_private_key_path: EDIT_HERE
You can spin up mesos box on ec2 by the same way with the case of virtual box
cd standalone
vagrant up --provider=aws
After box is up, you can see services running at:
http://#_public_dns_of_the_VM_#:5050
http://#_public_dns_of_the_VM_#:8080
http://#_public_dns_of_the_VM_#:8081
Tips: you can get public dns of the vm by:
$ vagrant ssh -- 'echo http://`curl --silent http://169.254.169.254/latest/meta-data/public-hostname`:5050'
http://ec2-54-193-24-154.us-west-1.compute.amazonaws.com:5050
Cluster configuration is defined at multinodes/cluster.yml
. You can edit the file to configure cluster settings.
# Mesos cluster configurations
mesos_version: 0.22.1
# The numbers of servers
##############################
zk_n: 1 # hostname will be zk1, zk2, …
master_n: 1 # hostname will be master1,master2,…
slave_n : 1 # hostname will be slave1,slave2,…
# Memory and Cpus setting(only for virtualbox)
##########################################
zk_mem : 256
zk_cpus : 1
master_mem : 256
master_cpus: 1
slave_mem : 512
slave_cpus : 2
# private ip bases
# When ec2, this should be matched with
# private addresses defined by subnet_id below.
################################################
zk_ipbase : "172.31.0."
master_ipbase: "172.31.1."
slave_ipbase : "172.31.2."
This takes several minutes(10 to 20 min.). It's time to go grabbing some coffee.
$ cd multinodes
$ vagrant up
At default setting, after all the boxes are up, you can see services running at:
this operations all VM instances forming the cluster.
$ cd multinodes
$ vagrant destroy
Because we assign private IP addreses to VM instances, this Vagrantfile requires Amazon VPC (you'll have to set subnet_id and security grooups both of which associates to the same VPC instance).
Note: Using default VPC is highly recommended. If you used non-default VPC, you should make sure to activate "DNS resolution" and "DNS hostname" feature in the VPC.
You have to configure some additional stuffs in multinodes/cluster.yml
which are related to EC2. Please note that
subnet_id
should be a VPC subnetsecurity_groups
should be ones associated to the VPC instance.
security_groups
should allow accesses to ports 22(SSH), 2181(zookeeper) and 5050--(mesos).(cont.)
# EC2 Configurations
# please choose one region from
# ["ap-northeast-1", "ap-southeast-1", "eu-west-1", "sa-east-1",
# "us-east-1", "us-west-1", "ap-southeast-2", "us-west-2"]
# NOTE: if you used non-default vpc, you should make sure that
# limit of the elastic ips is no less than (zk_n + master_n + slave_n).
# In EC2, the limit default is 5.
########################
access_key_id: EDIT_HERE
secret_access_key: EDIT_HERE
default_vpc: true # default vpc or not.
subnet_id: EDIT_HERE # VPC subnet id
security_groups: ["EDIT_HERE"] # array of VPN security groups. e.g. ['sg*** ']
keypair_name: EDIT_HERE
ssh_private_key_path: EDIT_HERE
region: EDIT_HERE
# see http://aws.amazon.com/ec2/instance-types/#selecting-instance-types
zk_instance_type: m1.small
master_instance_type: m1.small
slave_instance_type: m1.small
After editing configuration is done, you can just hit regular command.
$ cd multinode
$ vagrant up --provider=aws --no-parallel
NOTE: --no-parallel
is highly recommended because vagrant-berkshelf plugin is prone to failure in parallel provisioning.
After instances are all up, you can see
http://#_public_dns_of_the_master_N_#:5050
http://#_public_dns_of_marathon_#:8080
http://#_public_dns_of_chronos#:8081
if everything went well.
Tips: you can get public dns of the vms by:
$ vagrant ssh master1 -- 'echo http://`curl --silent http://169.254.169.254/latest/meta-data/public-hostname`:5050'
http://ec2-54-193-24-154.us-west-1.compute.amazonaws.com:5050
If you wanted to make sure that the specific mastar(e.g. master1
) could be an initial leader, you can cotrol the order of spinning up VMs like below.
$ cd multinode
# spin up an zookeeper ensemble
$ vagrant up --provider=aws /zk/
# spin up master1. master1 will be an initial leader
$ vagrant up --provider=aws master1
# spin up remained masters
$ vagrant up --provider=aws /master[2-9]/
# spin up slaves
$ vagrant up --provider=aws /slave/
# spin up marathon
$ vagrant up --provider=aws marathon
$ cd multinodes
$ vagrant halt
$ cd multinodes
$ vagrant reload --provision
This operations terminates all VMs instances forming the cluster.
$ cd multinodes
$ vagrant destroy