Stop using oslo.service graceful_shutdown_timeout

Add option in ironic.conf's condutor section:
`[conductor]/graceful_shutdown_timeout`.

Set `deprecated_group=DEFAULT` and add `deprecated_reason`.

Change-Id: I7dcc66802d0dc2baef7a932458ae255bfa0f9514

doc/source/admin/upgrade-guide.rst

@@ -185,7 +185,7 @@ older code, and start up a service using newer code with minimal impact.
 
 Nodes that are being acted upon by an ironic-conductor process, which are not in
 a stable state, will be put into a failed state when
-:oslo.config:option:`DEFAULT.graceful_shutdown_timeout` is reached. Node failures that occur
+:oslo.config:option:`conductor.graceful_shutdown_timeout` is reached. Node failures that occur
 during an upgrade are likely due to timeouts, resulting from delays involving
 messages being processed and acted upon by a conductor during long running,
 multi-step processes such as deployment or cleaning.
@@ -197,7 +197,7 @@ A drain shutdown is similar to graceful shutdown, differing in the following way
 
 * Triggered by sending signal ``SIGUSR2`` to the process instead of ``SIGTERM``
 * The timeout for process termination is determined by
-  :oslo.config:option:`DEFAULT.drain_shutdown_timeout` instead of :oslo.config:option:`DEFAULT.graceful_shutdown_timeout`
+  :oslo.config:option:`DEFAULT.drain_shutdown_timeout` instead of :oslo.config:option:`conductor.graceful_shutdown_timeout`
 
 :oslo.config:option:`DEFAULT.drain_shutdown_timeout` is set long enough so that any node in a not
 stable state will have time to reach a stable state (complete or failed) before

ironic/conductor/rpc_service.py

@@ -107,7 +107,7 @@ class RPCService(rpc_service.BaseRPCService):
         if self.draining:
             shutdown_timeout = CONF.drain_shutdown_timeout
         else:
-            shutdown_timeout = CONF.graceful_shutdown_timeout
+            shutdown_timeout = CONF.conductor.graceful_shutdown_timeout
         if shutdown_timeout == 0:
             # No timeout, run until no nodes are reserved
             return False

ironic/conf/conductor.py

@@ -607,6 +607,14 @@ opts = [
                     'here are validated as absolute paths and will be rejected'
                     'if they contain path traversal mechanisms, such as "..".'
                 )),
+    cfg.IntOpt('graceful_shutdown_timeout',
+               deprecated_group='DEFAULT',
+               deprecated_reason=_(
+                   'This replaces oslo.service '
+                   '[DEFAULT]/graceful_shutdown_timeout option.'),
+               default=60,
+               help='Specify a timeout after which a gracefully shutdown '
+                    'conductor will exit. Zero value means endless wait.'),
     cfg.DictOpt('bootloader_by_arch',
                 default={},
                 help=_(

releasenotes/notes/conductor-graceful-shutdown-timeout-db26a9194cf27bc2.yaml

@@ -0,0 +1,16 @@
+---
+upgrade:
+  - |
+    The configuration option `[DEFAULT]/graceful_shutdown_timeout` from
+    `oslo.service` has been replaced by configuration option
+    `[conductor]/graceful_shutdown_timeout`.
+
+    The default value (`60` seconds) has not changed, operators using a custom
+    value for should update their configuration files to use the new option.
+deprecations:
+  - |
+    The configuration option `[DEFAULT]/graceful_shutdown_timeout` from
+    `oslo.service` has been deprecated, replaced by
+    `[conductor]/graceful_shutdown_timeout`. This change was triggered by the
+    need to remove usage of eventlet.
+