How to have each node get a different variable from an array

Archives Ansible Project
1

Hi,

I am creating a playbook that creates a valid cassandra.yaml file for each node in a new cluster using a jinja2 template. What this file needs is a different token for each node, but I do not know how to do this with Ansible.

So the part of the template in question looks like this:

initial_token: {{ token }}

In this example, I have 6 nodes. I can generate the tokens for each using this script:

$ ./tokenbuilder.py 6
{
“0”: {
“0”: 0,
“1”: 28356863910078205288614550619314017621,
“2”: 56713727820156410577229101238628035242,
“3”: 85070591730234615865843651857942052864,
“4”: 113427455640312821154458202477256070485,
“5”: 141784319550391026443072753096570088106
}
}

I believe there is a way to get the output of that script into a variable in Ansible so that I can access it by {{ tokens[0] }}. (Any tips appreciated).

In any case, I can use a playbook that has

vars:

tokens:

0: 0

1: 28356863910078205288614550619314017621

2: 56713727820156410577229101238628035242

…

tasks:

  • name: Copy over the templated cassandra.yaml config file
    template: src=templates/cassandra.yaml.j2 dest=/etc/cassandra/cassandra.yaml

to do the same thing.

The question is, how can I then get each of the 6 hosts to use one of these array values in the template? So that when the template is evaluated:

host1 has token = tokens[0] = 0
host2 has token = tokens[1] = 28356863910078205288614550619314017621
host3 has token = tokens[2] = 56713727820156410577229101238628035242
host4 has token = tokens[3] = 85070591730234615865843651857942052864
host5 has token = tokens[4] = 113427455640312821154458202477256070485

host6 has token = tokens[5] = 141784319550391026443072753096570088106

All hosts are part of the same group in the inventory. Is there a trick to get the index of a host in a group?

I would really appreciate anyone pointing me in the right direction.

Thanks,
Peter

2

you can call a task that calls the script and outputs a token per
host, register the variable and then use that in the template.

3

Perhaps, but how?

How does ansible know which “host number in the group” it is running on?

This isn’t serial, all 6 hosts have the ansible playbook running simultaneously, so it is not a matter of maintaining a stack between each host where I pop off each token for each host.

To be clear, the tokens aren’t random, they are calculated based on the number of hosts (in this example, 6).

4

Hey Pete,

I am an newbie, so there might be better ways to do it, how about set the vars as
vars:
: 1234
: 456456
etc…

in in your template referece it as
token = {{ ansible_hostname }}

or maybe
set a custom facter in each host ‘hostnumber’ and then in template evaluate this value and set token accordingly

Regards,
Benno Joy

5

I like Benno's solution of the hash table, though you could also try
something like this to derive the token from the hostname:

- shell: echo $hostname | md5sum -t
  register: host_token

And then you could just loop over the hosts in your template::

{% for host in groups['groupname'] %}
    {{ hostvars[host]['host_token'] }}
{% endfor %}

6

Thanks for the suggestions. I believe both of those will work.

However, both suggestions push variables for one template in one playbook far up the stack and into the inventory. I am looking for a way to encapsulate all of this inside just the playbook itself without extra dependencies.

The other issue is that I don’t have named hosts (I am using the AWS EC2 inventory plugin). The hostnames are not unique and unpredictable.

To have a solution that is both encapsulated, and dynamic, I have come up with some Ansible/Jinja trickery.

Using this yaml file to store the pre-computed tokens for clusters of size 1 to 10 nodes: https://gist.github.com/pas256/5111146

I can use this in the my cassandra.yaml.j2 template:

Cassandra storage config YAML

{% set seeds = %}

{% for host in groups[‘tag_Name_cassandra’] -%}

{# This is in an if block even though it is just a variable assignment because

the only other way to append to an array is with the ‘do’ Jinja2 extension

-#}

{% if seeds.append(hostvars[host][‘ansible_eth0’][‘ipv4’][‘address’]) -%}{% endif -%}

{% endfor -%}

{% set num_hosts = seeds|length -%}

…

{% for ip in seeds|sort -%}

{% if ansible_eth0[“ipv4”][“address”] == ip -%}

initial_token: {{ initial_tokens[num_hosts][loop.index0] }}

{% endif -%}

{% endfor -%}

What I am doing is building an array of all nodes in the group, where the array entry is the ip address. Then further down the template, I can loop though that array looking for the IP address of the hosts that is running the template, and use the loop index as the index into the initial_tokens variable. To ensure consistency, the array is always sorted.

Not sure if there is a simpler way of doing this, but is does solve both my concerns, and has all the logic in the 1 place.

7

Thanks for sharing this… just saved me quite a bit of frustration.

I was trying something along these lines (https://groups.google.com/d/msg/ansible-project/YTF6Up3kaKw/xHASvMROhegJ) to get a list of private IPs for a MySQL cluster config, and have ended up with:

8

Hello,

Replying to an old thread but this is relevant to what I am trying to do.

I have defined the tokens as a dictionary in the vars_file as below.

tokens:
192.168.56.1: 0
192.168.56.2: 56713727820156410577229101238628035242
192.168.56.3: 113427455640312821154458202477256070485

192.168.56.x hosts are coming in from inventory file
and in the cassandra.yaml I have the following
initial_token: {{ tokens[‘{{ inventory_hostname}}’] }}

tokens[“{{ inventory_hostname}}”] is evaluated but not {{ tokens[‘{{ inventory_hostname}}’] }}

How do I get it done with jinja2? Where am I going wrong? Any help is much appreciated.

-SR