ControlMaster settings not taking affect

Archives Ansible Project
1

Hi,

I’ve noticed that Ansible isn’t using ControlMaster in my local setup, making it slower than it needs to be.

This is Ansible 1.6.1, running on OS X 10.9.3 and connecting to remote hosts running CentOS 6.

If I connect to the remote host and tail -f /var/log/secure I can see tell-tale Accepted publickey for ec2-user entries getting added over and over again during Ansible runs.

I tried overriding the ssh_connection ssh_args setting to specify a specific ControlMaster path as suggested in the docs, but the only thing that works is setting up a specific entry in ~/.ssh/config for the hosts I’m connecting to with (something like):

Host *

ControlMaster auto
ControlPath ~/.ssh/%r@%h:%p
ControlPersist 240

Any ideas why not this explicit set-up in ansible.cfg has any effect?

[ssh_connection]
ssh_args = -vvvv -o ControlMaster=auto -o ControlPersist=60s

control_path = %(directory)s/%%h-%%r

2

You should not need to specify those extra ssh_args if you’re using the ssh transport and your system supports ControlPersist. Can you gist/pastebin the output of a test run using -vvvvv to see any additional SSH debugging information available?

3

Alas, the problem went away. On running ansible-playbook with -vvvv, I could see the right Control* options getting used, and /var/log/secure no longer showed any re-auths.

Sorry for the noise. :frowning:

-Greg

4

No problem, let us know if it reoccurs.

Thanks!