Read a prediction as a prediction
Check asks Ansible for a dry run. The predicted change count is useful evidence, but it cannot tell you everything Apply will do. A check with no predicted changes is not a guarantee that an Apply will change nothing.
Some steps, including commands that do not support check mode, are skipped. Others depend on an earlier change that the dry run has not actually made. Read the notes about steps that cannot be predicted alongside the changes Ansible does predict.
Some steps can run for real
Some tasks can make real changes even during a dry run. Review your playbooks before pressing Check. A task, block or whole play can explicitly turn check mode off.
A step that ran for real during Check may do more on Apply. A command can also behave differently depending on whether it is a dry run. BareTold can point out these limits; it cannot predict the command’s effects for you.
Hidden details are another limit. If Ansible hides why a step was skipped, do not assume that check mode was the reason. BareTold cannot recover evidence Ansible withheld.
Before you choose Apply
- Review the predicted changes and any steps that could not be predicted.
- Read the notes for steps that run for real during Check or may do more on Apply.
- Consider whether the playbook, its inputs or the hosts have changed since the check. Check again if you need fresher evidence.
- Choose Apply when you are ready for a real run. It runs the playbook again, and its results can differ from Check.
For the first-run steps, see the quick-start help. Read why Apply can differ before relying on the predicted count.