Displaying a custom name for a host

Question

I have an Ansible play-book for working with EC2 instances. I'm using dynamic inventory (ec2.py) to get the group of instances that I want to work with (hosts: tag_Service_Foo). When I run it, it produces output like:

GATHERING FACTS ***************************************************************
ok: [54.149.9.198]
ok: [52.11.22.29]
ok: [52.11.0.3]

However, I can fetch the "Name" tag for a particular instance from Amazon (I do this and store it in a variable for use in a couple parts of the playbook).

Is there a way to get Ansible to use this string for the hostname when displaying progress? I'd like to see something more descriptive (since I don't have the IPs memorized):

GATHERING FACTS ***************************************************************
ok: [main-server]
ok: [extra-server]
ok: [my-cool-server]

The output of the ec2.py inventory script looks like this (truncated; it's very long).

{
  "_meta": {
    "hostvars": {
      "54.149.9.198": {
        "ec2__in_monitoring_element": false,
        "ec2_ami_launch_index": "0",
        "ec2_architecture": "x86_64",
        "ec2_client_token": "xxx",
        "ec2_dns_name": "xxx",
        "ec2_ebs_optimized": false,
        "ec2_eventsSet": "",
        "ec2_group_name": "",
        "ec2_hypervisor": "xen",
        "ec2_id": "i-xxx",
        "ec2_image_id": "ami-xxx",
        "ec2_instance_type": "xxx",
        "ec2_ip_address": "xxx",
        "ec2_item": "",
        "ec2_kernel": "",
        "ec2_key_name": "xxx",
        "ec2_launch_time": "xxx",
        "ec2_monitored": xxx,
        "ec2_monitoring": "",
        "ec2_monitoring_state": "xxx",
        "ec2_persistent": false,
        "ec2_placement": "xxx",
        "ec2_platform": "",
        "ec2_previous_state": "",
        "ec2_previous_state_code": 0,
        "ec2_private_dns_name": "xxx",
        "ec2_private_ip_address": "xxx",
        "ec2_public_dns_name": "xxx",
        "ec2_ramdisk": "",
        "ec2_reason": "",
        "ec2_region": "xxx",
        "ec2_requester_id": "",
        "ec2_root_device_name": "/dev/xvda",
        "ec2_root_device_type": "ebs",
        "ec2_security_group_ids": "xxx",
        "ec2_security_group_names": "xxx",
        "ec2_sourceDestCheck": "true",
        "ec2_spot_instance_request_id": "",
        "ec2_state": "running",
        "ec2_state_code": 16,
        "ec2_state_reason": "",
        "ec2_subnet_id": "subnet-xxx",
        "ec2_tag_Name": "main-server",
        "ec2_tag_aws_autoscaling_groupName": "xxx",
        "ec2_virtualization_type": "hvm",
        "ec2_vpc_id": "vpc-xxx"
      }
    }
  }
  "tag_Service_Foo": [
    "54.149.9.198",
    "52.11.22.29",
    "52.11.0.3"
  ],
}

Answer 1

For a more recent version, working since version 2.9 of Ansible, at least, the aws_ec2 plugin now allows this as a simple configuration, without the need to create any warper around ec2.py.

This can be done using a combination of the parameters hostnames and compose.

So the trick is to change the name of the EC2 instance via the hostnames parameter to have something human readable — for example from an AWS tag on the instance — and then to use the compose parameter to set the ansible_host to the private_ip_address, public_ip_address, private_dns_name or the public_dns_name to allow the connection, still.

Here would be the minimal configuration of the plugin to allow this:

plugin: aws_ec2
hostnames:
  - tag:Name 
  # ^-- Given that you indeed have a tag named "Name" on your EC2 instances
compose:
  ansible_host: public_dns_address

---

Mind, that, in AWS, the tags are not uniques, so you can have multiple instances tagged with the same name, but in Ansible, the hostname is something unique. This means that you will probably be better of composing the name with the tag:Name postfixing with something truly unique, like the instance ID.

Something like:

plugin: aws_ec2
hostnames:
  - name: 'instance-id'
    separator: '_'
    prefix: 'tag:Name'
compose:
  ansible_host: public_dns_address

Answer 2

If you put

vpc_destination_variable = Name

in your ec2.ini file, that should work too.

Answer 3

What you need to do is create your own wrapper (say my_ec2.py) over the ec2.py that would post process the output. Idea is to use the behavioral hostvar ansible_ssh_host. You can use any language not only python. As long as it prints valid json on stdout you're good to go. Reference if needed.

It'll be a tiny bit of work. But hope the sudo code would help:

output_json_map = new map
for each group in <ec2_output>: # e.g. tag_Service_Foo, I think there would be another 
                                # key in the output that contains list of group names.
  for each ip_address in group:
    hname = ec2_output._meta.hostvars.find(ip_address).find(ec2_tag_Name)

    # Add new host to the group member list
    output_json_map.add(key=group, value=hname)
    copy all vars from ec2_output._meta.hostvars.<ip_address>
                  to output_json_map._meta.hostvars.<hname>
    # Assign the IP address of this host to the ansible_ssh_host
    # in hostvars for this host
    output_json_map.add(key=_meta.hostvars.<hname>.ansible_ssh_host,
                        value=ip_address)
    output_json_map.add(key=_meta.hostvars.find(ip_address).ansible_ssh_host,
                        value=ip_address)

print output_json_map to stdout

E.g. for your example the output of my_ec2.py should be:

{
  "_meta": {
    "hostvars": {
      "main-server": {
        "ansible_ssh_host": "54.149.9.198"
        --- snip ---
        "ec2_tag_Name": "main-server",
        --- snip ---
      },
      "extra-server": {
        "ansible_ssh_host": "52.11.22.29"
        --- snip ---
        "ec2_tag_Name": "extra-server",
        --- snip ---
      },
      <other hosts from all groups>
    }
  }
  "tag_Service_Foo": [
    "main-server",
    "extra-server",
    <other hosts in this group>
  ],
  "some other group": [
    <hosts in this group>,
    ...
  ],
}

and obviously, use this my_ec2.py instead of ec2.py as the inventory file. :-)

-- edit --

1) In the groups, can I only refer to things by one name? 2) There's no notion of an alias? 3) I'm wondering if I could use the IP addr in the groups and just modify the _meta part or if I need to do it all?

Yes*, No and no.

---

* Technically first yes should be no. Let me explain.

What we are doing here can be done with static inventory file like this:

Original ec2.py was returning json equivalent of following inventory file:

[tag_Service_Foo]
54.149.9.198 ec2_tag_Name="main-server"  ec2_previous_state_code="0" ...
52.11.22.29  ec2_tag_Name="extra-server" ec2_previous_state_code="0" ...

our new my_ec2.py returns this:

[tag_Service_Foo]
main-server  ansible_ssh_host="54.149.9.198" ec2_tag_Name="main-server"  ec2_previous_state_code="0" ...
extra-server ansible_ssh_host="52.11.22.29"  ec2_tag_Name="extra-server" ec2_previous_state_code="0" ...

# Technically it's possible to create "an alias" for main-server like this:
main-server-alias  ansible_ssh_host="54.149.9.198" ec2_tag_Name="main-server"  ec2_previous_state_code="0" ...

Now you would be able to run a play with main-server-alias in the host list and ansible would execute it on 54.149.9.198.

BUT, and this is a big BUT, when you run a play with 'all' as the host pattern ansible would run the task on main-server-alias as well as main-server. So what you created is an alias in one context and a new host in another. I've not tested this BUT part so do come back and correct me if you find out otherwise.

HTH