When Ansible receives a non-zero return code from a command or a failure from a module, by default it stops executing on that host and continues on other hosts. However, in some circumstances you may want different behavior. Sometimes a non-zero return code indicates success. Sometimes you want a failure on one host to stop execution on all hosts. Ansible provides tools and settings to handle these situations and help you get the behavior, output, and reporting you want.
By default Ansible stops executing tasks on a host when a task fails on that host. You can use ignore_errors
to continue on in spite of the failure:
- name: Do not count this as a failure ansible.builtin.command: /bin/false ignore_errors: yes
The ignore_errors
directive only works when the task is able to run and returns a value of ‘failed’. It does not make Ansible ignore undefined variable errors, connection failures, execution issues (for example, missing packages), or syntax errors.
New in version 2.7.
You can ignore a task failure due to the host instance being ‘UNREACHABLE’ with the ignore_unreachable
keyword. Ansible ignores the task errors, but continues to execute future tasks against the unreachable host. For example, at the task level:
- name: This executes, fails, and the failure is ignored ansible.builtin.command: /bin/true ignore_unreachable: yes - name: This executes, fails, and ends the play for this host ansible.builtin.command: /bin/true
And at the playbook level:
- hosts: all ignore_unreachable: yes tasks: - name: This executes, fails, and the failure is ignored ansible.builtin.command: /bin/true - name: This executes, fails, and ends the play for this host ansible.builtin.command: /bin/true ignore_unreachable: no
If Ansible cannot connect to a host, it marks that host as ‘UNREACHABLE’ and removes it from the list of active hosts for the run. You can use meta: clear_host_errors
to reactivate all hosts, so subsequent tasks can try to reach them again.
Ansible runs handlers at the end of each play. If a task notifies a handler but another task fails later in the play, by default the handler does not run on that host, which may leave the host in an unexpected state. For example, a task could update a configuration file and notify a handler to restart some service. If a task later in the same play fails, the configuration file might be changed but the service will not be restarted.
You can change this behavior with the --force-handlers
command-line option, by including force_handlers: True
in a play, or by adding force_handlers = True
to ansible.cfg. When handlers are forced, Ansible will run all notified handlers on all hosts, even hosts with failed tasks. (Note that certain errors could still prevent the handler from running, such as a host becoming unreachable.)
Ansible lets you define what “failure” means in each task using the failed_when
conditional. As with all conditionals in Ansible, lists of multiple failed_when
conditions are joined with an implicit and
, meaning the task only fails when all conditions are met. If you want to trigger a failure when any of the conditions is met, you must define the conditions in a string with an explicit or
operator.
You may check for failure by searching for a word or phrase in the output of a command:
- name: Fail task when the command error output prints FAILED ansible.builtin.command: /usr/bin/example-command -x -y -z register: command_result failed_when: "'FAILED' in command_result.stderr"
or based on the return code:
- name: Fail task when both files are identical ansible.builtin.raw: diff foo/file1 bar/file2 register: diff_cmd failed_when: diff_cmd.rc == 0 or diff_cmd.rc >= 2
You can also combine multiple conditions for failure. This task will fail if both conditions are true:
- name: Check if a file exists in temp and fail task if it does ansible.builtin.command: ls /tmp/this_should_not_be_here register: result failed_when: - result.rc == 0 - '"No such" not in result.stdout'
If you want the task to fail when only one condition is satisfied, change the failed_when
definition to:
failed_when: result.rc == 0 or "No such" not in result.stdout
If you have too many conditions to fit neatly into one line, you can split it into a multi-line yaml value with >
:
- name: example of many failed_when conditions with OR ansible.builtin.shell: "./myBinary" register: ret failed_when: > ("No such file or directory" in ret.stdout) or (ret.stderr != '') or (ret.rc == 10)
Ansible lets you define when a particular task has “changed” a remote node using the changed_when
conditional. This lets you determine, based on return codes or output, whether a change should be reported in Ansible statistics and whether a handler should be triggered or not. As with all conditionals in Ansible, lists of multiple changed_when
conditions are joined with an implicit and
, meaning the task only reports a change when all conditions are met. If you want to report a change when any of the conditions is met, you must define the conditions in a string with an explicit or
operator. For example:
tasks: - name: Report 'changed' when the return code is not equal to 2 ansible.builtin.shell: /usr/bin/billybass --mode="take me to the river" register: bass_result changed_when: "bass_result.rc != 2" - name: This will never report 'changed' status ansible.builtin.shell: wall 'beep' changed_when: False
You can also combine multiple conditions to override “changed” result:
- name: Combine multiple conditions to override 'changed' result ansible.builtin.command: /bin/fake_command register: result ignore_errors: True changed_when: - '"ERROR" in result.stderr' - result.rc == 2
See Defining failure for more conditional syntax examples.
The command and shell modules care about return codes, so if you have a command whose successful exit code is not zero, you can do this:
tasks: - name: Run this command and ignore the result ansible.builtin.shell: /usr/bin/somecommand || /bin/true
Sometimes you want a failure on a single host, or failures on a certain percentage of hosts, to abort the entire play on all hosts. You can stop play execution after the first failure happens with any_errors_fatal
. For finer-grained control, you can use max_fail_percentage
to abort the run after a given percentage of hosts has failed.
If you set any_errors_fatal
and a task returns an error, Ansible finishes the fatal task on all hosts in the current batch, then stops executing the play on all hosts. Subsequent tasks and plays are not executed. You can recover from fatal errors by adding a rescue section to the block. You can set any_errors_fatal
at the play or block level:
- hosts: somehosts any_errors_fatal: true roles: - myrole - hosts: somehosts tasks: - block: - include_tasks: mytasks.yml any_errors_fatal: true
You can use this feature when all tasks must be 100% successful to continue playbook execution. For example, if you run a service on machines in multiple data centers with load balancers to pass traffic from users to the service, you want all load balancers to be disabled before you stop the service for maintenance. To ensure that any failure in the task that disables the load balancers will stop all other tasks:
--- - hosts: load_balancers_dc_a any_errors_fatal: true tasks: - name: Shut down datacenter 'A' ansible.builtin.command: /usr/bin/disable-dc - hosts: frontends_dc_a tasks: - name: Stop service ansible.builtin.command: /usr/bin/stop-software - name: Update software ansible.builtin.command: /usr/bin/upgrade-software - hosts: load_balancers_dc_a tasks: - name: Start datacenter 'A' ansible.builtin.command: /usr/bin/enable-dc
In this example Ansible starts the software upgrade on the front ends only if all of the load balancers are successfully disabled.
By default, Ansible continues to execute tasks as long as there are hosts that have not yet failed. In some situations, such as when executing a rolling update, you may want to abort the play when a certain threshold of failures has been reached. To achieve this, you can set a maximum failure percentage on a play:
--- - hosts: webservers max_fail_percentage: 30 serial: 10
The max_fail_percentage
setting applies to each batch when you use it with serial. In the example above, if more than 3 of the 10 servers in the first (or any) batch of servers failed, the rest of the play would be aborted.
Note
The percentage set must be exceeded, not equaled. For example, if serial were set to 4 and you wanted the task to abort the play when 2 of the systems failed, set the max_fail_percentage at 49 rather than 50.
You can also use blocks to define responses to task errors. This approach is similar to exception handling in many programming languages. See Handling errors with blocks for details and examples.
See also
An introduction to playbooks
Tips and tricks for playbooks
Conditional statements in playbooks
All about variables
Have a question? Stop by the google group!
#ansible IRC chat channel
© 2012–2018 Michael DeHaan
© 2018–2021 Red Hat, Inc.
Licensed under the GNU General Public License version 3.
https://docs.ansible.com/ansible/latest/user_guide/playbooks_error_handling.html