Using register with exact_count when launching an instance on EC2

[Ofer Herman]

Hi,

I'm using exact_count when  creating an instance on EC2, when the instance already exists the next action fails with the following error:

fatal: [localhost] => One or more undefined variables: 'unicode object' has no attribute 'public_ip'

This is the playbook:

- name: Provision EC2 SC instances

  local_action:

        module: ec2            

        region: "{{ aws_region }}"

        keypair: "{{ aws_keypair }}"

        group: ansibleDefault

        instance_type: "{{ aws_instanceType }}"

        image: "{{ aws_image }}"

        exact_count: 1

        count_tag: 

          Class: "websever"

        vpc_subnet_id: "{{ vpc.subnets[0].id }}"

        assign_public_ip: True

        instance_tags: { "Class":"webserver" }

        wait: yes

  register: ec2_info

- name: Wait for SSH to come up

  local_action: wait_for host={{ item.public_ip }} port=22 delay=60 timeout=320 state=started

  with_items: ec2_info.instances

How can the playbook continue even if the instance exist?

Thanks,

Ofer

[Michael Peters]

Try using debug on ec2_info to see what it contains. It seems that
instances is coming back with a list of strings and not a list of
objects like you thought.

> --
> You received this message because you are subscribed to the Google Groups
> "Ansible Project" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to ansible-proje...@googlegroups.com.
> To post to this group, send email to ansible...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/ansible-project/1f0d1a69-9bf6-4021-9d4f-33b53d907d4b%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Michael DeHaan]

For clarity, this:

- debug: var=ec2_info

To view this discussion on the web visit https://groups.google.com/d/msgid/ansible-project/CAJQqANfcEk%2BvanSap0oEw93y-g_-_mxOQo4p-N%2BpO-0Lco64UQ%40mail.gmail.com.

[Stuart Reynolds]

ec2_info.tagged_instances? Does that give all of the instances matching the tag, whether or not they we created in the prior step. 
Test with:

- debug: var=ec2_info

[Craig Marvelley]

Was a solution ever found for this problem? Have the same problem where sometimes the ec2 module launches instances and sometimes it doesn't, but I want to work with the data of the running instances regardless.