Merge pull request #14838 from resmo/patch-3

Proposal: re-run-handlers

docs/proposals/re-run-handlers.md

@@ -0,0 +1,77 @@
+# Proposal: Re-run handlers cli option
+
+*Author*: René Moser <@resmo>
+
+*Date*: 07/03/2016
+
+- Status: New
+
+## Motivation
+
+The most annoying thing users face using ansible in production is running handlers manually after a task failed after a notified handler.
+
+### Problems
+
+Handler notifications get lost after a task failed and there is no help from ansible to catch up the notified handlers in a next ansible playbook run.
+
+~~~yaml
+- hosts: localhost
+  gather_facts: no
+  tasks:
+    - name: simple task
+      shell: echo foo
+      notify: get msg out
+
+    - name: this tasks fails
+      fail: msg="something went wrong"
+
+  handlers:
+    - name: get msg out
+      shell: echo handler run
+~~~
+
+Result:
+
+~~~
+$ ansible-playbook test.yml
+
+PLAY ***************************************************************************
+
+TASK [simple task] *************************************************************
+changed: [localhost]
+
+TASK [this tasks fails] ********************************************************
+fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg": "something went wrong"}
+
+NO MORE HOSTS LEFT *************************************************************
+
+RUNNING HANDLER [get msg out] **************************************************
+	to retry, use: --limit @test.retry
+
+PLAY RECAP *********************************************************************
+localhost                  : ok=1    changed=1    unreachable=0    failed=1
+~~~
+
+## Solution proposal
+
+Similar to retry, ansible should provide a way to manully invoke a list of handlers additionaly to the notified handlers in the plays:
+
+~~~
+ $ ansible-playbook test.yml --notify-handlers <handler>,<handler>,<handler>
+ $ ansible-playbook test.yml --notify-handlers @test.handlers
+~~~
+
+Example:
+
+~~~
+ $ ansible-playbook test.yml --notify-handlers "get msg out"
+~~~
+
+The stdout of a failed play should provide an example how to run notified handlers in the next run:
+
+~~~
+...
+RUNNING HANDLER [get msg out] **************************************************
+	to retry, use: --limit @test.retry --notify-handlers @test.handlers
+~~~
+