Forks when hosts are specified in playbooks

[Mzp45]

Hello,

I have a playbook where the hosts are listed by name instead of "all".  

---

- hosts: host1

  roles:

     - role1

     - role2

- hosts: host2

     - role 1

     - role 3

     - role 5

and so on

However this doesn't help with the forking as it will only execute one host at a time. Is there a way to set it to fork when the playbook is arranged out in this manner?

Thanks

[David Karban]

Hi,

that is how ansible works, plays are run in order and paralelism is on play level. I would consider to make several groups in your inventory:

---

- hosts: group1

  roles:

    - role1

- hosts: group2

  roles:

    - role2

- hosts: group3

  roles:

    - role3

    - role5

this way at least role1 will be parallelized.

David

--
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/462fb88e-f33a-4f95-8916-648b371dae34%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

David Karban
Specialista na správu linuxových serverů
www.karban.eu

[Mzp45]

Hi Dave,

Thanks for the input. Yeah I kind of know about that but we are not using groups (we do have a reason for that). I'll keep looking around to see if i can get a solution for this particular problem

Thanks

[Michael DeHaan]

"we do have a reason for that" -- what's the reason?

Even having groups of one to designate a role can be a good idea.

To view this discussion on the web visit https://groups.google.com/d/msgid/ansible-project/f4013f7c-f093-4db3-9ce6-9e6addcdac7c%40googlegroups.com.

[Mzp45]

A lot of our customers are wary about running roles/updates/fixes across multiple hosts. They prefer having granular control and usually specify which systems they need work done on. But at the same time they'd like to keep track of what roles have been applied to each machine. 

Thanks

[Michael DeHaan]

This is possibly a good case for a staging environment or rolling updates, Ansible can be configured to fail if a certain % of a batch don't make it, before adding them back into a pool, etc.

--Michael

To view this discussion on the web visit https://groups.google.com/d/msgid/ansible-project/5dd6e4a1-2e1a-495c-8142-83b578c429f3%40googlegroups.com.