AWS user_data with Packer

旧街凉风 提交于 2020-01-01 08:53:33

问题


So I'm trying to use Packer to create an AWS image and specify some user data via user_data_file. The contents of this file needs to be run when the instance boots as it will be unique each time. I can't bake this into the AMI.

Using packer I have the following:

{
  "variables": {
  "ami_name": ""
  },
  "builders": [
  {
    "type": "amazon-ebs",
    "region": "us-east-1",
    "source_ami": "ami-c8580bdf",
    "instance_type": "t2.micro",
    "ssh_username": "ubuntu",
    "ami_name": "{{ user `ami_name` }}-{{ isotime | clean_ami_name }}",
    "user_data_file": "user_data.sh",
    "tags": {
      "os_version": "ubuntu",
      "built_by": "packer",
      "build_on": "{{ isotime | clean_ami_name }}",
      "Name": "{{ user `ami_name` }}"
    }
  }],
  "provisioners": [
  {
    "type": "ansible",
    "playbook_file": "playbook.yml",
    "user": "ubuntu"
  }]
}

The contents of my user_data shell script are just a few basic config lines for a package that was installed via the ansible scripts that were run in the provisioners step. Watching the output of Packer I can confirm that the ansible scripts all run.

Packer completes and creates the AMI, but the user data piece is never executed. No record of it exists in resulting image. There is no /userdata.log file and /var/lib/cloud/instance/user-data.txt is empty I feel like I missing something basic as this should be a very simple thing to do with Packer.


回答1:


Rereading this I think maybe you misunderstood how user-data scripts work with Packer.

user_data is provided when the EC2 instance is launched by Packer. This instance is in the end, after provisioning snapshoted and saved as an AMI.

When you launch new instances from the created AMI it doesn't have the same user-data, it gets the user-data you specify when launching this new instance.

The effect of the initial (defined in your template) user-data might or might not be present in the new instance depending if the change was persisted in the AMI.




回答2:


As pointed out by Rickard von Essen the answer was to copy my script to /var/lib/cloud/scripts/per-instance which would execute my script on every instance launched from this AMI.

Alternately you can put your script in /var/lib/cloud/scripts/per-boot if you needed this to happen each time the instance boots.

In my case since I wanted to register the instance with a 3rd party service I only had it execute once per instance creation.




回答3:


Uploading to /var/lib/cloud/scripts/* will work, but it depends on how you want your images built. Do you need to be able to spin up an instance quickly?

The best solution would be to us Packer provisioners. Provisioners are used to install and configure the machine image after booting by using ansible/salt/puppet/cheff/shell scripts etc, you can provision your image with what ever you need. That way you are not tied down to having to provision deps on each instance launch, which can cause some issues (think intermittent network issues/failures, which could cause some deps not to be installed)

The provisioners for packer are third party u



来源:https://stackoverflow.com/questions/45110795/aws-user-data-with-packer

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!