Is there any way to make an SSH connection error a fatal error, that stops the play?

[Duncan Lock]

Occasionally, I get SSH timeouts or other network errors in the middle of a play. These cause that task to fail, but not the whole playbook, which continues on, but without the node that failed. You see output something like this:

...

2014-11-24 23:54:08,717 p=21193 u=ubuntu |  TASK: [common | Install latest pip from PyPi] ****************************

2014-11-24 23:54:12,034 p=21193 u=ubuntu |  changed: [box0]

2014-11-24 23:54:09,443 p=21193 u=ubuntu |  fatal: [box1] => SSH encountered an unknown error during the connection. We recommend you re-run the command using -vvvv, which will enable SSH debugging output to help diagnose the issue

2014-11-24 23:54:11,001 p=21193 u=ubuntu |  changed: [box2]

2014-11-24 23:54:11,278 p=21193 u=ubuntu |  changed: [box3]

2014-11-24 23:54:12,305 p=21193 u=ubuntu |  changed: [box4]

...

You eventually get this at the bottom of the output:

box0: ok=29   changed=23   unreachable=0    failed=0   

box1: ok=18   changed=15   unreachable=1    failed=0   

box2: ok=77   changed=61   unreachable=0    failed=0   

box3: ok=77   changed=62   unreachable=0    failed=0   

box4: ok=77   changed=62   unreachable=0    failed=0

This means that box1 doesn't get anything that should have happened after the task that failed, because it got dropped from the play - and all groups - at that point.

This means that I occasionally get one cluster node that doesn't work properly, because it was only partially provisioned/setup.

Is there a way to make this SSH error fatal and to stop the whole playbook/ansible run at that point - the same way an error in a task itself would?

[Brian Coca]

check this setting, I believe it will allow you to do what you want
http://docs.ansible.com/playbooks_delegation.html#maximum-failure-percentage
--
Brian Coca

[Duncan Lock]

Thanks!