One of my recent task was to restart a service after a database migration. The "problem" was that this service was deployed using aws autoscaling group, and so there was close to 10 instances. This post is about how to automate this.

Grab all private IPs for a given autoscaling group

First, the aws cli can be used to get a big json blog describing all running ec2 instances.

aws ec2 describe-instances help

There is a way to narrow down the result list using filters. The list is huge, and one looked promising: image-id. All instances in an autoscaling group use the same AMI (amazon vm images), so that's easy to use.

aws ec2 describe-instances --filters Name=image-id,Values=ami-xxxx

The result is a big json blob, and I'm only interested in the private IPs of the instances. To get these easily, I used jq, a command-line JSON processor. The syntax is a bit weird at first, but it's a very powerful and convenient tool, always handy to have.

jq ".Reservations[].Instances[0].PrivateIpAddress"

This retrieve all items in the toplevel Reservations object (it's an array). Then, It get the first Instances item, and focus on the PrivateIpAddress

The final command is:

aws ec2 describe-instances --filters Name=image-id,Values=ami-xxxx \
| jq ".Reservations[].Instances[0].PrivateIpAddress" > ips

Execute a command for all instances

Good old bash is perfect for this job:

for ip in $(cat ips); do ssh username@${ip:1:-1} 'ls -al'; done

Note the use of substrings. The ips retrieved in the previous section are quoted, so we need to remove them with the syntax ${ip:1:-1} which acts the same as python ranges: my_string[1:-1]