Hi,
I have a dynamic inventory of ec2 hosts that have tags to declare how those should be provisioned. My use-case is as follows:
- 1 EC2 host has a tag “tomcat” and “use”
- 2 EC2 hosts have a tag “mysql”
one of them has a tag “use”
on of them has a tag “dont-use”
I’d like to provision only the “use” hosts. The challenge is that the tomcat host needs a variable with the ip address of the mysql host. I currently have a playbook that matches on all hosts in the “use” group. Then I apply the roles using when: ‘mysql’ in group_names, ‘tomcat’ respectively.
roles:
- {
role: ‘my_tomcat_role’,
mysql_ip: “WHAT TO PUT HERE?”,
when: “‘tomcat’ in group_names”,
sudo: yes
}
I have played around with various techniques. Eg. try to dynamically group_by in the pre tasks to create temp groups of all hosts that have been matched by the playbook but that does not work, because group_by seems to implicitly only run_once and thus can not use the group_names:
- group_by: key=current_{{ item }}
with_items: “{{ group_names }}”
The above does not work.
I can not register variables because those will only be available per host, so that does not work either because I don’t know the hostname of the mysql-tagged instance.
Any ideas? Maybe this is a feature request to allow to select hostvars/hosts by the same pattern syntax that is available in the playbook hosts: directive.
Thanks,
Nicolas