EC2 Cluster Auto-discovery for GO

This package provides simple cluster auto-discovery for go using EC2 metadata. This should be used to seed a more reliable group membership algorithm, such as github.com/hashicorp/memberlist. Most such algorithms require an initial seed list of nodes, which is what ec2cluster provides.

In this example we use all the EC2 instances tagged with the same value for app as the initial known members for a memberlist cluster:

ec2TagName := "app"
knownMembers, err := DiscoverClusterMembersByTag(&aws.Config{}, ec2TagName)
if err != nil {
    panic(err)
}
cluster := memberlist.Create(memberlist.DefaultLANConfig())
_, err := cluster.Join(knownMembers)
if err != nil {
    panic(err)
}
defer cluster.Leave(time.Second)

When run as a command this program produces output that can be used to set environment variables. You can invoke it like:

eval $(ec2cluster)

The output it produces can be passed to the shell, or parsed:

AVAILABILITY_ZONE="us-west-2a"
REGION="us-west-2"
ADVERTISE_ADDRESS="10.0.10.124"
TAG_AWS_CLOUDFORMATION_STACK_ID="arn:aws:cloudformation:us-west-2:012345678901:stack/myapp/d0a9d5eb-dfed-4455-ab4b-5e7214f337bb"
TAG_APPLICATION="myapp"
TAG_AWS_AUTOSCALING_GROUPNAME="example-Cluster1-Q0YWRWQJC5XL"
TAG_CLUSTER="myapp-cluster1"
TAG_AWS_CLOUDFORMATION_STACK_NAME="myapp"
TAG_AWS_CLOUDFORMATION_LOGICAL_ID="Cluster1"
TAG_NAME="myapp.example.com-master"
ASG_DESIRED_CAPACITY=3
ASG_MIN_SIZE=1
ASG_MAX_SIZE=5
CLUSTER="10.0.10.124 10.0.188.207 10.0.118.6"

Monitoring ASG Lifecycle events

You can also use this tool to monitor autoscaling lifecycle events. To do this, configure your autoscaling group with a lifecycle hook that emits events to an SQS queue. Then invoke ec2cluster --watch which will produce one line of output per event, like:

autoscaling:EC2_INSTANCE_TERMINATING    i-403e6d87

crewjam/ec2cluster

EC2 Cluster Auto-discovery for GO

Monitoring ASG Lifecycle events