Few tips for troubleshooting Ansible, based on my brief experiences with Ansible 1.4 (read: do not rely on this info too much).

Run ansible-playbook in the verbose mode

ansible-playbook -vvvv ... will provide you with plenty of details of what is going on. (Notice that additional v:s, starting from none, add more detail.)

Use ./hacking/test-module

Check out Ansible sources and use the ./hacking/test-module script - see Developing Modules.

Keep remote files and execute them manually

Note: This is quite a hack; it is preferred to use the the test-module.

Set the environment variable ANSIBLE_KEEP_REMOTE_FILES=1 to keep files that Ansible copied to the server so that you can execute them directly on the server yourself and thus troubleshoot them better:

$ export ANSIBLE_KEEP_REMOTE_FILES=1
$ ansible-playbook ...

Then, on the server, in the home directory of the user used for ssh-ing, look into .ansible/tmp/ansible-<unique id>/<module name or "arguments">. F.ex. to execute manually a command that has failed:

$ python ./ansible-1392625678.65-85585787027596/command

Common issues

The halting issue: Ansible "freezes" while executing a command

The command is perhaps waiting for user input. Always make sure to execute commands in a non-interactive mode if possible.

Try to find the command process on the server and kill it while running Ansible with -vvvv or/and use ANSIBLE_KEEP_REMOTE_FILESand run the Ansible files manually, as suggested above.

Ansible fails right at start, perhaps with "Authentication or permission failure", for a new node

Ansible currently doesn't handle SSH problems very transparently, even when run with -vvvv. Such a failure could be caused by SSH asking to accept the node's key. In each case, exclude SSH as the cause by SSH-ing into the node manually in the same way as Ansible does so when you run Ansible, it will already have been added to known_hosts.

Note: I believe this is fixed/improved in Ansible 1.5.

Tags: DevOps