isabella232 / vsi-ci-cd

vsi-ci-cd

This is a companion repository to the blog post: Design and Build a Delivery Pipeline for Virtual Server Images

Create an image using packer. Demonstrate the roll out in two different environments:

1. simple - vpc with one instance
2. autoscale - instance group with autoscale configuration

Prepare

Prerequisites

Permissions to create resources, including VPC, instances, etc.

• VPC SSH key
• Use the IBM Cloud Shell or make sure you have Shell with ibmcloud cli, terraform, jq and git

Packer must be installed. In the IBM Cloud Shell you can do the following:

wget https://releases.hashicorp.com/packer/1.8.3/packer_1.8.3_linux_amd64.zip
unzip packer_1.8.3_linux_amd64.zip
PATH=`pwd`:$PATH

Configure

Get the source and edit the local.env file

git clone https://github.com/IBM-Cloud/vsi-ci-cd
cd vsi-ci-cd
cp template.local.env local.env

Use your editor to make the changes described in the comments:

edit local.env

Terraform 1.2.8 or better is required. The tfswitch tool will update if installed (it is available on IBM Cloud Shell). If you do not have this install terraform:

(cd image_tf && tfswitch)

See make clean below to remove up all the resources.

make

Use make to invoke each of the steps that you may want to add to your pipeline, make x will execute one of the example steps

make vpc

Invoke terraform in the image_tf directory to create a vpc and subnet for packer. The terraform output captures the packer variables.

make vpc

make image

Invoke Packer to create an image. The image created is recorded in packer-manifest.json see packer docs for details. This step is Image Pipeline in the diagram.

make image

make tag

Delete the tag from all images and then add the tag to the latest packer image. The file packer-manifest.json will be used to identify the latest image.

make tag

make simple

Invoke terraform in the simple_tf directory to create a vpc, subnet and instance with the image just tagged in make tag.

make simple

Test simple via curl

Find the public ip address of the simple instance and curl:

public_ip=$(terraform output -state simple_tf/terraform.tfstate -raw public_ip)
echo $public_ip
while sleep 1; do curl $public_ip; done

or:

make simple_curl

make autoscale

The autoscale production make target contains two steps, both executed in the vpc_autoscale_tf directory:

• apply.sh - wrapper around terraform apply that juggles two instance_template resources.
• roll.sh - Slowly delete each of the members of the instance_group allowing them to be replaced by the latest packer image

make autoscale

Test autoscale via curl

Find the public DNS name of the load balancer and curl:

load_balancer=$(terraform output -state vpc_autoscale_tf/terraform.tfstate -raw load_balancer_hostname)
echo $load_balancer
while sleep 1; do curl $load_balancer; done

or:

make autoscale_curl

make rollout

Now that everythine is working you can make a change to the image and roll out a new version. The contents of the image are defined in ubuntu-hello.pkr.hcl look for these sections:

locals {
  timestamp = regex_replace(timestamp(), "[- TZ:]", "")
  version = local.timestamp
  ...
}

  provisioner "shell" {
    inline = [
      "set -x",
      "echo '@reboot echo ${local.version} $(hostname) $(hostname -I) > /var/www/html/index.html' | crontab",
      ...

The echo is putting a line into the crontab file that is executed on reboot to write some text to index.html. Evaluating hostname and hostname -I on the instance during boot provides some info about the VPC instance. The ${local.version} is evaluated by packer and is derived from the timestamp at the time packer was run. Add some code to the echo line like this:

  provisioner "shell" {
    inline = [
      "set -x",
      "echo '@reboot echo FUN ${local.version} $(hostname) $(hostname -I) > /var/www/html/index.html' | crontab",
      ...

Steps to roll out a new image:

make image
make tag
make simple
make autoscale

This will create a new image and provision. Test via curl described above. You can transfer the public_ip and load_balancer variables to your laptop computer to watch as the new images are rolled out.

make clean

Clean up all of the resources.